Skip to main content

Full text of "soil analysis"

See other formats


- 



tB6NOM4-HH 



Soil salinity assessment 

Methods and interpretation 

of electrical conductivity measurements 



' ») /-^ 







IRRIGATION 
AND DRAINAGE 



UtliMJ 





Food 

and 

Ayr iu_il turu 

Organisation 

of 

the 

United 

Nations 




(fWv) 







Soil salinity assessment 

Methods and interpretation 

of electrical conductivity measurements 



by 

J.D. Rhoades 

US Salinity Laboratory 

United States Department of Agriculture 

Riverside, California, USA 

R Chanduvi 

Water Resources, Development and Management Service 

FAO Land and Water Development Division 

S. Lesch 

US Salinity Laboratory 

United States Department of Agriculture 

Riverside, California, USA 









57 










■ 












Food 

and 

Agriculture 

Orqanizalion 

of 

the 

United 

Nations 






™S| 




Rome. 1 999 





The designations employed and the presentation of material in this 
Replication do not imply the expression of any opinion whatsoever on 
the part of the Food and Agriculture Organization of the United 
Nations concerning the legal status of any country, territory, city or 
area or of its authorities, or concerning the delimitation of its frontiers 
or boundaries. 



M-51 
ISBN 92-5-104281-0 



All rights reserved. No part of this publication may be reproduced, stored in a 
retrieval system, or transmitted in any form or by any means, electronic, mechani- 
cal, photocopying or otherwise, without the prior permission of the copyright owner. 
Applications for such permission, with a statement of the purpose and extent of the 
reproduction, should be addressed to the Director, Information Division, Food and 
Agriculture Organization of the United Nations, Viale delle Terme di Caracalla, 
00100 Rome, Italy. 

© FAO 1999 



Soil salinity assessment Mi 



Foreword 



The technology described in this report for measuring soil salinity has been extensively and 
successfully field-tested. It is concluded to be sound, reliable, accurate and applicable to a wide 
variety of useful applications. It is based on proven theory of soil electrical conductivity. The 
required equipment is commercially available. The advocated instrumental methodology is 
practical, cost effective and well developed for essentially all general applications. It is cheaper, 
faster and more informative than traditional methods of salinity measurement based on soil 
sampling and laboratory analyses. Software is available to facilitate its use for mapping and 
monitoring uses, as is equipment to mobilize and automate the measurements for use in detailed 
field-scale assessments. Its usefulness has been demonstrated: 1) for diagnosing soil salinity, 2) 
for inventorying soil salinity, 3) for monitoring soil salinity, 4) for evaluating the adequacy and 
appropriateness of irrigation and drainage systems and management practices, 5) for determining 
the areal sources of excessive leaching, drainage and salt-loading in crop lands, 6) for 
establishing the spatial soil information needed to develop prescription farming plans to manage 
fields with spatially-variable salinity conditions, and 7) for scheduling and controlling irrigations 
under saline conditions. It offers the potential to identify the inherent causes of salinization in 
fields, especially when integrated with GIS technology, and to identify mitigation needs, 
especially when integrated with field-scale deterministic, solute-transport models. The salinity 
assessment approach advocated in this report offers a more suitable basis for evaluating, 
managing and controlling soil salinity than do the leaching requirement and salt balance 
concepts/measurements as traditionally applied. National programs need to be implemented to 
mitigate the substantial problems of secondary salinization that threaten the sustainability of 
irrigation in many places in the world. Holistic; meaningful salinity assessment approaches 
needed in this regard are illustrated in this report. The presented salinity assessment technology 
offers substantial practical potential to inventory, monitor, manage and control soil and water 
salinity, as will be needed to sustain irrigated agriculture and to meet the worlds food needs in the 
coming decades. 



IV 



Acknowledgements 



The authors wish to acknowledge the valuable advice provided by the staff of the Land and Water 
Development Division, who revised and contributed to the general outline of the publication. The 
technical reviews of the first draft by Messrs N.K. Tyagi (Central Soil Salinity Research 
Institute, Karnal, India) and J.W. van Alphen (International Agricultural Centre, The 
Netherlands) are gratefully appreciated. 

Most of the research findings and recommendations in this publication come from the United 
States Salinity Laboratory, Riverside, California, and the authors acknowledge the Laboratory 
staff for their valuable contribution. 

The authors with to express their gratitude to Dr Arumugan Kandiah, Programme Manage of the 
International Programme for Technology and Research in Irrigation and Drainage (IPTRID) of 
FAO for his technical and overall support towards the fulfilment of this publication. 

Thanks are also due to Ms CD. Smith Redfern for her valuable assistance in the preparation of 
the final camera-ready text and figures and tables. She also contributed to the improved 
presentation of the document. 



Soil salinity assessment 



Contents 



Page 

1 INTRODUCTION 1 

2 DETERMINATION OF SOIL SALINITY FROM AQUEOUS ELECTRICAL 5 
CONDUCTIVITY 

Principles of aqueous electrical conductivity 5 

Soil water salinity 6 

Soil extract salinity 1 1 

3 DETERMINATION OF SOIL SALINITY FROM SOIL-PASTE AND BULK SOIL 15 
ELECTRICAL CONDUCTIVITY 

Principles of soil and soil-paste electrical conductivities 15 

Determining soil salinity from saturated soil-paste electrical conductivity 29 

Determining soil salinity from bulk soil electrical conductivity 33 

Sensors and equipment for measuring soil electrical conductivity 33 

Procedures for measuring bulk soil electrical conductivity 39 

Procedures for interpreting soil salinity 48 

Comparisons of the different methods of measuring soil salinity 58 

Determination of locations of measurement and calibration sites 6 1 

4. EXAMPLE USES OF SALINITY ASSESSMENT TECHNOLOGY 63 

Diagnosis of soil salinity and saline seeps 63 

Inventorying soil salinity 66 

Monitoring soil salinity 69 

Developing information for site-specific management 70 

Evaluating adequacy and appropriateness of irrigation/drainage 76 

Assessing leaching and salt-loading 84 

Scheduling and controlling irrigations 91 

Reclamation of saline soils 93 

5 OPERATIONAL AND EQUIPMENT COSTS ASSOCIATED WITH SALINITY 95 
INSTRUMENTATION MEASUREMENT TECHNIQUES 

Salinity instrumentation: equipment specifications and cost information 96 

Soil salinity survey instruments 96 

GPS equipment 98 

Mobilized soil salinity survey systems 99 

Analytical (laboratory) conductivity instruments 102 

Soil salinity assessment and mapping software 102 

Company information 103 



VI 



Page 

Operational costs associated with the appraisal of soil salinity 103 

Operational costs associated with conventional sampling 105 

Operational costs associated with survey instrumentation 107 

Conventional sampling versus survey instrumentation costs in multi- 111 
field (large area) survey applications 

Cost advantages associated with instrument mobilization 113 

Conclusion 117 

REFERENCES 119 

ANNEX 1 Methods for establishing ECe = F(ECa) calibrations 129 

ANNEX 2 Circuitry and parts-list for soil EC-meter 133 

ANNEX 3 Equation for calculating effect of insertion-depth of four-electrodes 135 

ANNEX 4 Construction of burial-type four-electrode probe 137 

ANNEX 5 Examples of various special-use four-electrode cells and sensors 139 

ANNEX 6 Derivation of EM a dj equations 143 

ANNEX 7 Device for positioning EM-38 sensor during hand-held measurements 145 

ANNEX 8 Schematic and parts-list for soil four-electrode probe 147 

ANNEX 9 Description of statistical tests for monitoring soil salinity 149 



Soil salinity assessment vii 



List of figures 



Page 

1 Diagram of vacuum extractor apparatus for sampling soil water 8 

2 (A) Imbibition type salinity sensor with spring, housing, and pin in disassembly; 

(B) schematic of internal elements of salinity sensor 9 

3 Commercial meter and salinity sensor showing ceramic disc in which platinum 
electrodes are imbedded 10 

4 Variations in soil water electrical conductivity (EC W ) and soil water tension in the 

root zone of an alfalfa crop during the spring of the year 1 1 

5 (A) Saturated soil-paste in mixing container and in drying/weighing container, 
(B) saturated soil-paste being extracted by vacuum, and (C) collection of 
saturated-paste extract 12 

6 Measurement of the electrical conductivity of a saturated-paste extract (EG) 

using a laboratory micro-conductivity cell 13 

7 Model and schematic representation of electrical conduction in soil: (A) the three 
paths that current can take in unsaturated soil; (B) simplified soil electrical 
conductance model consisting of the three conductance elements (a-c) acting in parallel 16 

8 Electrical conductivity of Waukena loam soil as a function of the electrical 
conductivity and volumetric content of soil water 1 8 

9 Correlation between EG and clay percentage for a number of soils from the 

S an Joaquin Valley of C alif ornia 1 9 

10 Relationship between the volumetric content of soil water existing in the series- 
path (@ws), the continuous-path (@» , c ), and the total water content (@») found 

for various California soils 20 

1 1 Relationship between the product of soil water electrical conductivity (ECw ) 
and volumetric water content (@ w ) and the product of the electrical conductivity 
of the saturated-paste extract (EC e ), its saturation percentage (SP), and the 
soil bulk density (pb ) found for soils of the Wellton- Mohawk Irrigation 

and Drainage District, Arizona, USA 21 

12 Correlation between slopes of EC e vs. Ed calibrations obtained for different 

soils and their saturation percentage (SP) and bulk densities (pb) 22 

13 Correlation between slopes of EC e vs. EG, calibrations obtained for different 

soils and saturation percentages 22 

14 Relationship found between saturation percentage and clay percentage for 

some California soils 23 

15 Relationship between slopes of EC e vs. EG calibrations obtained for 

different soils of Montana USA and their clay contents 23 

16 Correlation between slopes of EC e vs. EG calibrations obtained for 

different soils and their field-capacity water contents 24 



VIII 



Page 

17 (A) Cylinder and surrounding "moat" with impounded saline water used to leach 
the soil and adjust it to a desired level of salinity; (B) access-hole being made in 
soil with Oakfield-type soil sampling tube for subsequent insertion of Ed-probe; 
(C) ECa-probe being inserted into salinity- adjusted soil for determination of Ed; 
and (D) sample of salinized soil being collected for subsequent determination 

of ECe (salinity) 25 

1 8 Soil-filled, four-electrode cell (as obtained with a coring device) showing one 
group of four of the eight electrodes inserted into the undisturbed soil used 

to measure EG, ; after the soil is removed, it is analyzed in the laboratory for EC e 26 

19 Relationship between bulk soil electrical conductivity (Ed) and the electrical 
conductivity of the saturated-paste extract (EG, soil salinity) for Dateland 

soil at field capacity water content 26 

20 Relationships between bulk soil electrical conductivity and soil water 
electrical conductivity for the major soils of the Welfton-Mohawk 

Irrigation Project of Arizona USA 27 

21 Relationship between the soil electrical conductivity of Fallbrook soil and the 

electrical conductivity and sodium adsorption ratio (SAR) of the soil water 27 

22 Response of soil electrical conductivity to irrigations and evapotranspiration 

over several irrigation cycles 28 

23 Theoretical relation between saturation percentage (SP) and weight (in grams) 
of 50 cm of saturated paste, assuming a particle density of 2.65 g/cm 

24 Correspondence between measured and estimated (using Figure 23) saturation 
percentages, for a set of California soils 29 

25 Correlation between the electrical conductivity of the soil solid phase (EG) 
and the saturation percentage (SP) for some soils of the San Joaquin Valley 

of California USA 30 

26 Relationship between the volumetric content of water in the saturated soil-paste 
which is in the continuous electrical conduction path (@ w - @ ws ) and the 

saturation percentage (SP) for some San Joaquin Valley California USA soils 30 

27 (A) Portable balance used in the field to determine the weight of the saturated 
soil-paste filling the "Bureau of Soils Cup", (B) "Bureau of Soils Cup" 
filled with saturated soil-paste connected to conductance meter, and 

(C) close up of "Bureau of Soils Cup" 31 

28 Relationships between EC P , EG and SP for representative arid- land soils 32 

29 Relationships between EC P and EG for Grangeville soil 32 

30 Correspondence between measured and estimated (using Figure 29) soil 
salinities (EG) for representative soils of the San Joaquin Valley of 

California USA 33 

31 Four electrodes positioned in a Wenner-type surface- array and a combination 

electrical generator and resistance meter 34 



Soil salinity assessment ix 



Page 

32 Schematic showing increased depth and volume of EG measurement with increased 
Ci -C2 electrode spacing. Effective depth of measurement is approximately equal to 
one-third of (Ci - C2). C stands for current-electrode and P stands for potential- 
measuring electrode 34 

33 Variation of: (A) current density with depth in a plane mid-way between the 
current electrodes; (B) current density at unit depth as a function of current 

electrode separation 35 

34 A " fixed-array" four-electrode apparatus and commercial generator/meter 35 

35 A mobilized (tractor-mounted) "fixed-array" four-electrode system, with mast 

for GPS antenna 36 

36 (A) Close-up of the four fixed-array electrodes used on the mobilized tractor- 
mounted system, (B) insulators used to isolate sensor part of shank from the rest 

of the tractor, and (C) close up of replaceable pad at bottom of electrode 36 

37 Two commercial four-electrode probes (small and standard sizes) and electrical 
generator-meter/data-logger 37 

38 Commercial burial-type four-electrode conductivity probe used for monitoring 

changes in soil electrical conductivity 37 

39 Schematic showing the principle of operation of an electromagnetic induction 

soil conductivity sensor 38 

40 Geonics EM-38 electromagnetic soil conductivity sensor in (A) horizontal orientation 

and (B) vertical orientation 38 

41 Mobilized salinity assessment system with combined EM-38 and four-electrode 

soil conductivity sensors and GPS antenna 38 

42 Close-up of the fixed-array four-electrode unit in (A) the travel-position and 
(B) inserted into the soil by the hydraulic "scissors" apparatus of the 

mobilized combination sensor assessment system 39 

43 Schematic of distances between the current and potential electrodes in four-electrode 

array for use with Equation [19] 40 

44 Cumulative relative contribution of all soil electrical conductivity, R(Z), below 
various depths sensed by the EM-38 unit when placed on the soil surface in 

horizontal (parallel) and vertical (perpendicular) magnetic-coil positions 42 

45 Ratio of vertical and horizontal weighted responses of the EM-38 unit as a function 

of composite depth increments (i.e., 0-0.15, 0- 0.30, 0-0.45, 0-0.60 m, etc.) 43 

46 EM-38 readings for homogeneous profiles as a function of the profile EG value 43 

47 Theoretical relation between In EMh and the difference (In EMh - In EMv ) for 

uniform Ed profiles 45 

48 Relationship between soil electrical conductivity (EG), as determined with 
interelectrode spacings and measured average soil salinity (EG) for a glacial- till 

soil in Montana USA 49 

49 Relationships between soil electrical conductivity (EG ) and salinity (expressed as 

EC e ) for representative soil types of the northern Great Plains USA 49 



Page 

50 Relationship between EC X , as calculated from Equation [25], and soil salinity 
(expressed as EC e ), for soil-depth intervals of 0-30, 30-60, 60-90, and 90-120 cm 

for a glacial- till soil in Montana USA 52 

5 1 Graphs of measured and calculated (by three different methods) EG -depth profiles 

for three California USA sites 52 

52 Relationships between electrical conductivity of bulk soil (Ed), electrical conductivity 
of saturated-paste extract (ECe), relative soil water content as percent of field-capacity, 

and soil clay content (% clay), for representative arid- land soils 54 

53 Relationship between electrical conductivity of bulk soil (EG) and " Wenner-array" 
interelectrode spacing (approximately equivalent to soil depth) for a saline seep, 

an encroaching saline seep site, and an unaffected site for glacial-till soil in 

Montana USA 64 

54 Relationship between electrical conductivity of bulk soil (EG) in the 0-30 cm 
depth- increment and depth to water table in typical glacial-till soils of 

Montana USA 64 

55 Maps showing:(A) surface topography and location of saline seep, and marginally 
and unaffected alfalfa crop surrounding it; (B) isolines of EG in the 0-30 cm. soil 
depth-interval, and (C) isolines of EG in the 30-60 cm. soil depth-interval 65 

56 Antenna, battery pack and meter for measuring location on the landscape 

using the LORAN technique 66 

57 Contour map of: (A) measured and (B) predicted soil salinities, in 15-square 

mile study area in Central California USA 67 

58 Maps of measurement and sampling locations and of measured and predicted soil 
salinity patterns in a field (Hanford S2A) located in the San Joaquin Valley of 
California USA 68 

59 Three-dimensional map of the electrical conductivity of bulk soil (EG) of a salt- 
affected field in the Coachella Valley of California USA 71 

60 Estimated salinity (ECe in dS/m) of the 0-60 cm depth-interval of a salt-affected 

field in the Coachella Valley of California USA 72 

61 Estimated sodicity (SARe basis) of the 0-60 cm depth-interval of a salt- affected 

soil in the Coachella Valley of California USA 72 

62 Estimated salinity/sodicity ratio of the 0-60 cm depth-interval of a salt-affected 

field in the Coachella Valley of California USA 73 

63 Three-dimensional map of the electrical conductivity of bulk soil (EG) of a 
salt-affected, Sudangrass field in the Coachella Valley of California USA 73 

64 Relationship between the electrical conductivity of bulk soil (EG) and distance 
across (and perpendicular to the sub-surface drainage system) a salt-affected 
Sudangrass field in the Coachella Valley of California USA 74 

65 Relationship between the EM-38 sensor readings of EG , Sudangrass plant height, 
and distance across (and perpendicular to the sub-surface drainage system) a 
salt-affected sudangrass field in the Coachella Valley of California USA 75 



Soil salinity assessment xi 



Page 

66 Three-dimensional map of the predicted Sudangrass yield (in terms of height) predicted 
from the data of Figures 64 and 65 for a salt-affected Sudangrass field in the 
Coachella Valley of California USA 

67 Relationship between (a) soil electrical conductivity (EC a ), as measured by both the 
mobilized, four-electrode system and the mobilized, electromagnetic (EM) system, and 
(b) measured and predicted soil salinity (EC e basis) and distance along a transect across 

a furrow-irrigated, sugar beet field located in the Imperial Valley of California USA 76 

68 Correspondence between soil salinity predictions based on soil electrical conductivity 
measurements obtained with the mobile salinity assessment system along a transect 
across a surface irrigated, tile-drained alfalfa field (Imperial clay soil) located in the 
Imperial Valley of California 77 

69 Relationship between soil electrical conductivity (ECa) and distance along a transect 
crossing two sets of subsurface tile-drains in a salt-affected field located in the 
Coachella Valley of California USA 78 

70 Predicted average root zone soil salinity in a tile-drained field located in the 

Coachella Valley of California USA 79 

71 Relationship between salinity distribution and mean level of salinity in the soil profile 

of a tile-drained field located in the Coachella Valley of California USA 79 

72 Salinity distribution in the soil profiles of a tile-drained field located in the Coachella 
Valley of California USA, as influenced by mean (0-0.5 m) salinity level 80 

73 Average salinity distribution in the soil profiles along a transect across a furrow- 
irrigated, tile-drained alfalfa field (Imperial clay soil) located in the Imperial 

Valley of California USA 80 

74 Relationship between the salinity profile ratio and leaching fraction 82 

75 (A) Predicted soil salinity (EC e ) and (B) salinity profile ratio for a tile-drained field 
located in the Coachella Valley of California USA 82 

76 Cyclic pattern of soil electrical conductivity (ECa) across a succession of furrows, 

some of which were trafficked by a tractor (V) and some which were not 83 

77 Relationship between electrical conductivity of soil water and leaching fraction 
for the Colorado River water used in the Wellton- Mohawk Irrigation and Drainage 
District of Arizona, USA 88 

78 Calibrations established between leaching fraction (Lf) and electrical conductivity 
of soil (ECa) for different soil types in the Wellton-Mohawk Irrigation and Drainage 
District of Arizona, USA 89 

79 Correlation found between electrical conductivity of soil water (EC ) and the ratio 

of chloride concentration in soil water (Cl» ) below the rootzone to that in the Colorado 
River irrigation water for the Indio fine silty loam soil at four study sites in the 
Wellton-Mohawk Irrigation and Drainage District of Arizona, USA 89 

80 Water penetration following a small irrigation, as deduced from measurements made 

with four-electrode sensors and a neutron probe 92 

8 1 Water penetration following an irrigation of moderate amount, as deduced from 
measurements of soil electrical conductivity (ECa) with four-electrode sensors 92 

82 Plot of soil electrical conductivity vs. depth of water infiltrated during the ponded 
leaching of salinized Pachappa soil 94 



XII 



Page 

1.1 Photograph of a series of four-electrode cells containing undisturbed soil-cores 

being segmented after removal from the soil-core sampler 130 

1.2 Comparison of EC e -EC a calibrations as determined by methods (1) and (3) 131 

2.1 Diagram of simplified circuit or EC a -meter; R y , the resistance of the null adjustment 
potentiometer, is adjusted to equal R,, the resistance of the salinity sensor 133 

2.2 Detailed diagram of low-cost circuit for reading four-electrode sensors 134 
4.1 Inexpensive four-electrode burial- type probe before and after assembly 137 

5.1 Schematic of apparatus used to vary water content and composition and of a four- 
electrode cell used to measure the associated values of soil electrical conductivity 139 

5.2 Several kinds of four-electrode cells and bases used to vary water content and 
composition and to measure the associated values of soil electrical conductivity 140 

5.3 Schematic of apparatus used to hold four-electrode cells for studying the effects of 
varying solution composition and induced changes in water content and porosity 

and to measure the associated values of soil electrical conductivity 140 

5.4 An apparatus built to hold a four-electrode cell with ceramic containing end-plates 

to permit changes in water content to be induced by the application of pressure 141 

5.5 Small-span four-electrode sensor used to measure soil electrical conductivity in small 
depth-increments along an exposed soil profile 141 

5.6 (A) Small-span four-electrode array used to measure soil electrical conductivity in 
shallow depths, and (B) connection of meter to four-electrode cell 142 

6.1 Relationship between electromagnetic soil conductivity as measured by the EM-38 in 
the horizontal position at the soil surface, EMo, h (m easuredi and adjusted 
electromagnetic soil conductivity, EMo, h (adjusted) for composite depths 144 

7.1 EM-38 sensor (centre of coils) positioned 10 cm above ground in the (A) vertical 

and (B) horizontal orientations, respectively 145 

7.2 EM-38 sensor (centre of coils) positioned 50 cm above ground in the (A) vertical 

and (B) horizontal orientations, respectively 146 

7.3 Wooden device used to position EM-38 at 10- and 50-cm heights above ground 146 

8.1 Schematic of the design and parts list for the construction of a four-electrode 

soil EC -probe 147 

8.2 (A) An unassembled and (B) assembled four-electrode soil EC-probe 148 



Soil salinity assessment xiii 



Symbols 



e a random error component 

(3 a regression- fitted parameter estimate 

a depth-weighted mean anion exclusion volume as a proportion of the volumetric soil 
water content above depth z in the soil 

X the ratio of the solute concentration at depth z to the mean concentration; sjs a 

pb bulk density of the soil 

°C temperature in degrees Celsius 

@e volumetric content of soil water of a saturated paste 

@f C volumetric content of soil water at "field-capacity" 

©m mean volumetric content of soil water averaged over depth z 

p s density of the soil particles 

0s volumetric content of the solid phase of soil which does not contain an indurated layer 

@sc volumetric content of solid phase of soil (which contains an indurated layer) in a 
continuous electrical solid-pathway 

@ss volumetric content of solid phase of soil (which is contains an indurated layer) in a 
series-coupled electrical solid-pathway 

ASsw the change in soil solution salinity within the rootzone depth 

0» volumetric content of total soil water 

@» c volumetric content of soil water in the continuous-liquid pathway; essentially that in the 
larger, continuous pores ("mobile water") 

@» s volumetric content of soil water in the series-coupled pathway; essentially that in the 
small pores & films ("immobile water") 

b z proportion of applied water moving as by-pass flow past depth z in soil 

c concentration of tracer solute in the irrigation water 

Cdw concentration of salt in drainage water flowing from the rootzone 

C g w concentration of salt in shallow groundwater 

Gw concentration of salt in irrigation water 

Ct» concentration of surface runoff water (tailwater) 

EC2 5 electrical conductivity referenced to a temperature of 25 °C 

ECip ECa measured using four-electrode array 

ECa electrical conductivity of bulk soil 

ECa depth-weighted value of ECa 

EC e electrical conductivity of the extract of a saturated soil-paste 

EC S electrical conductivity of surface conductance of soils without indurated layer 

ECsc electrical conductivity of indurated solid phase 

ECss electrical conductivity of surface conductance of soils with indurated solid phase 



XIV 



ECt electrical conductivity at the temperature of the sample 

EC W c electrical conductivity of @ w c 

EC W s electrical conductivity of @ w s 

Ed EC a within a depth-interval, as estimated from a sequence of surface-array four- 
electrode readings 

EM electromagnetic induction 

EMh EM measurement made with the axis of the magnet-coil in the horizontal configuration 

EMv EM measurement made with the axis of the magnet-coil in the vertical configuration 

ft proportion of flow that occurs at the concentration of soil water 

I rate of irrigation averaged over period of measurement or calculation 

L leaching flux at depth z averaged over period of measurement or calculation 

Lf leaching fraction; Vd* /Vnf 

log the logarithm of a value to the base ten 

S c amount of salt removed from soil solution by crop uptake 

Sf amount of salt added to the soil solution by the dissolution of fertilizers and 
amendments 

S m amount of salt brought into soil solution by mineral weathering 

S m mean concentration of conservative solute averaged over depth z 

Sm (o ) mean concentration of conservative solute averaged over depth z at time to 

S P amount of salt removed from soil solution by the precipitation of salt-minerals 

SP saturation percentage; the gravimetric water content of a saturated soil-paste 

s z mean concentration of tracer solute at depth z averaged over period of measurement at 

the reference water content @ m 

T [temperature in degrees Celsius - 25] /10 

t time period of measurement or calculation 

to initial time of measurement, or reference time for beginning of calculation 

Vcu volume of water consumed by crop in evapo transpiration 

Vd» volume of drainage water flowing from the soil rootzone 

V g w volume of groundwater which flows into the rootzone depth of soil 

Vint volume of infiltrated water (V,» - V» ) 

V» volume of applied irrigation water 

V volume of net leaching (Vw - V« w ) 

Vw volume of surface runoff water (tailwater) 

W a trend surface matrix based on the spatial coordinates of the measurement sites 

X a matrix of log transformed and de-correlated sensor readings 

Y the vector of log transformed soil salinity values 
z depth in soil profile 



Soil salinity assessment 



Chapter 1 
Introduction 



The world's demand for food is increasing at such a rate that the ability to meet anticipated needs 
in the next several decades is becoming questionable. Irrigated agriculture presently accounts for 
about one- third of the world's production of food and fibre; it is anticipated that it will need to 
produce nearly 50 percent by the year 2040 (FAO, 1988). This will likely be difficult, because 
extensive areas of irrigated land have been and are increasingly becoming degraded by 
salinization and waterlogging resulting from over-irrigation and other forms of poor agricultural 
management (Ghassemi, et al., 1995). Available data suggest that the present rate of such 
degradation has surpassed the present rate of expansion in irrigation (Seckler, 1996). In some 
places, the very sustainability of irrigated agriculture is threatened by this degradation (Rhoades, 
1997a; Rhoades, 1998). At the same time, irrigated agriculture is also depleting and polluting 
water supplies in many places. Increased irrigation efficiency is being sought to conserve water, 
to reduce drainage, waterlogging and secondary salinization, and to mitigate some of the water 
pollution associated with irrigated agriculture. Restrictions are increasingly being placed on the 
discharge of saline drainage water from irrigation projects. Concomitantly, the reuse of saline 
drainage water for irrigation is being increased. With less leaching and drainage discharge and 
greater use of saline water for irrigation, soil salinity may increase in some areas. Thus, a 
practical methodology is needed for the timely assessment of soil salinity in irrigated fields, for 
determining its causes and for evaluating the appropriateness of related management practices. 

Ideally, it would be desirable to know the concentrations of the individual solutes in the soil 
water over the entire range of field water contents and to obtain this information immediately in 
the field. Practical methods are not available at present to permit such determinations, although 
determinations of total solute concentration (i.e., salinity) can be made in situ using electrical or 
electromagnetic signals from appropriate sensors. Such immediate determinations are so valuable 
for salinity diagnosis, inventorying, monitoring and irrigation management needs that, in many 
cases, they supplant the need for soil sampling and laboratory analyses. However, if knowledge 
of a particular solute(s) concentration is needed (such as when soil sodicity or the toxicity of a 
specific ion are to be assessed) then either a sample of soil, or of the soil water, is required to be 
analysed. Of course, the latter methods require much more time, expense and effort than the 
instrumental field methods. In this case, a combination of the various instrumental and laboratory 
methods should be used to minimize the need for sample collection and chemical analyses, 
especially when monitoring solute changes with time and characterizing the salinity conditions of 
extensive areas. 

Customarily, soil salinity has been defined and assessed in terms of laboratory- 
measurements of the electrical conductivity of the extract of a saturated soil-paste sample (EC e ; 
this as well as all other symbols used in this report are summarized in the list of symbols). This is 



Introduction 



because electrical conductivity is an easily measured and practical index of the total concentration 
of ionized solutes in an aqueous sample. The saturation percentage (SP) is the lowest water/soil 
ratio suitable for the practical laboratory extraction of readily dissolvable salts in soils (US 
Salinity Laboratory, 1954). As the water/soil ratio approaches that of a field soil, the 
concentration and composition of the extract approaches that of soil water. Soil salinity can also 
be determined from the measurement of the electrical conductivity of a soil-water sample (EC W ). 
This latter measurement can be made either in the laboratory on a collected sample or directly in 
the field using in situ, imbibition-type salinity sensors. Alternatively, salinity can be indirectly 
determined from measurement of the electrical conductivity of a saturated soil-paste (EC P ) or 
from the electrical conductivity of the bulk soil (EC a ). EC P can be measured either in the 
laboratory or in the field using simple and inexpensive equipment. EC a can be measured in the 
field either using electrical probes (electrodes) placed in contact with the soil or remotely using 
electromagnetic induction devices. The latter two sensors are more expensive than those used to 
measure the EC of water samples, of soil-extracts or soil-pastes. However, their use is very cost 
effective when one considers the amount of spatial information that can be acquired with them 
(the relative costs of the different methods for assessing soil salinity are discussed later; the basis 
for this economic evaluation is presented in Chapter 5). From measurements of EC P and EC a , soil 
salinity can be deduced in terms of either EC e or EC W . The appropriate sensor and method to use 
depends upon the purpose of the salinity determination, the size of the area being evaluated, the 
number and frequency of measurements needed, the accuracy required and the available 
equipment/human resources. 

Traditionally, the leaching requirement (L,) and salt-balance-index (SBI) concepts have 
been used to judge the appropriateness of irrigation and drainage systems and practices, with 
respect to the avoidance of salinity and waterlogging problems; these concepts have also been 
used to estimate the extra water requirements associated with saline irrigated lands (US Salinity 
Laboratory Staff, 1954). However, these approaches are either inadequate or impractical for 
these purposes. The leaching requirement (L,), refers to the amount of leaching required to 
prevent excessive loss in crop yield caused by salinity build-up within the root zone from the 
salts applied in the irrigation water. Its calculation is based on the assumption of steady-state 
and of uniform conditions of irrigation, infiltration, leaching and evapotranspiration; none of 
which are achieved in most field situations which typically are dynamic and variable, both 
spatially and temporally in the above mentioned attributes. Furthermore, salt build-up in the root 
zone resulting from the presence of shallow water tables is ignored in the traditional L 
calculation. Additionally, no practical way has existed to directly measure the degree of leaching 
actually being achieved in a given field, much less in the various parts of it, as is required in order 
to determine its appropriateness. However, a potential means has been developed to estimate the 
extent and adequacy of leaching based on measurements of the levels and distributions of salinity 
within irrigated root zones, as will be described later. 

The salt-balance index (SBI), which is the net difference between the amount of salt added 
to an irrigation project and that removed in its drainage effluent, is another "concept" that 
traditionally has been used to evaluate the appropriateness of irrigation, leaching and drainage 
practices. This approach is also inadequate for these purposes because it provides no information 
about the average level of soil salinity in the project, nor about the actual level of soil salinity 
existing within any specific field of the project. The approach also fails because it does not even 
provide a realistic measure of trends in salinity within the root zone, because salt derived from 
below the soil profile and of geologic origin is typically contained in the drainage water collected 
by the subsurface drainage system (Kaddah and Rhoades, 1976). Additionally, the transit times 



Soil salinity assessment 



involved in the drainage returns are so long (often more than 25 years) that the index values are 
not reflective of current conditions/trends (Jury, 1975a, 1975b). From project-wide SBI values, 
one can not deduce the extent of leaching being achieved in any field, nor the uniformity and 
efficiency of irrigation and leaching, nor the extent of waterlogging- and of salinity-induced losses 
in crop yield; traditional SBI measurements are impractical to make on the basis of individual 
fields and of root zone environments. 

In the author's opinion, the effective control of soil salinity and waterlogging, and also of 
salinity in drainage-receiving waters, requires the following: (i) knowledge of the magnitude, 
extent and distribution of root zone soil salinity in representative fields of the irrigation project (a 
suitable inventory of conditions); (ii) knowledge of the changes and trends of soil salinity over 
time and the ability to determine the impact of management changes upon these conditions (a 
suitable monitoring programme); (iii) ways to identify the existence of salinity problems and 
their causes, both natural and management-induced (a suitable means of detecting and 
diagnosing problems and identifying their causes); (iv) a means to evaluate the appropriateness 
of on-going irrigation and drainage systems and practices with respect to controlling soil salinity, 
conserving water and protecting water quality from excessive salinization (a suitable means of 
evaluating management practices), (v) an ability to determine the areas where excessive deep 
percolation is occurring, i.e., to identify the diffuse sources of over-irrigation and salt loading (a 
suitable means of determining areal sources of pollution), (vi) knowledge of the spatial 
variability in soil salinity needed to develop site-specific management "tailored" to deal with such 
variability and to avoid excessive and wasteful inputs of irrigation, fertilizer and other potentially 
harmful and costly cropping inputs (a suitable means of establishing the spatial-variability of 
soil salinity at the field scale), and (vii) a methodology for including soil salinity in the 
determination of plant-available soil water and for guiding irrigation management (a suitable 
means for scheduling and controlling irrigations under saline conditions). 

From measurements of the levels and distributions of soil salinity within the root zones of 
individual fields, one can determine whether, or not, salinity is within acceptable limits for crop 
production. One can also infer from these measurements whether, or not, leaching and drainage 
are adequate anywhere in a field, since soil salinity is a tracer of the net processes of infiltration, 
leaching, evapotranspiration and drainage. Thus, a more appropriate and practical approach for 
assessing the adequacy of salinity control than either the Lr or SBI approaches is the acquisition 
of periodic, detailed information of soil salinity levels and distributions within the root zones of 
representative individual fields of the project. The same data can also be used for delineating the 
sources of salt- loading in fields and irrigated landscapes, as well as for mapping the distribution 
and extent of drainage problem areas, both at the project and field scales. The author refers to the 
above described approach as "salinity assessment" and advocates its use to diagnose, inventory 
and monitor soil salinity, as well as to evaluate the appropriateness of leaching and drainage and 
to guide management practices. 

An assessment technology of the type described above begins with a practical methodology 
for measuring soil salinity in the field. This is complicated by the spatially variable and dynamic 
nature of soil salinity, which is caused by the effects and interactions of varying edaphic factors 
(soil permeability, water table depth, salinity of perched groundwater, topography, soil parent 
material, geohydrology), by management-induced factors (irrigation, drainage, tillage, cropping 
practices), as well as by climate-related factors (rainfall, amount and distribution, temperature, 
relative humidity, wind). Numerous samples (measurements) are needed to characterize just one 
field and the measurements often need to be updated as conditions change, or to determine if they 
are changing. When the need for extensive sampling requirements and repeated measurements are 



Introduction 



met, the expenditure of time and effort to characterize and monitor the salinity condition of a 
large area with conventional soil sampling and laboratory analysis procedures becomes 
impractical (as is shown later). Soil salinity is too variable and transient to be appraised using the 
numbers of samples that can be practically processed using conventional soil sampling and 
laboratory analysis procedures. Furthermore, the conventional procedures do not provide 
sufficient detailed spatial information to adequately characterize salinity conditions nor to 
determine its natural or management-related causes. A more rapid, field-measurement technology 
is needed. Additionally, this assessment technology should ascertain the spatial relations existing 
within extensive areal data sets. It should also provide a means for evaluating management effects 
and for proving changes or differences in an area salinity condition over time. 

A system of the type advocated above has been developed. It consists of mobile 
instrumental techniques for rapidly measuring bulk soil electrical conductivity (EC a ) directly in 
the field as a function of spatial position on the landscape, procedures and software for inferring 
salinity from EC a , computer-assisted mapping techniques capable of associating and analysing 
large spatial databases, and appropriate spatial statistics to infer salinity distributions in root 
zones and to detect changes in salinity over space and time. It will be described in some detail in 
this report. The complementary use of geographic information systems and remote sensing 
technology to facilitate the determination of the underlying causes of the observed salinity 
conditions would extend the utility of this system. The additional use of solute transport models 
utilizing the spatial data provided by the assessment system, as a basis to predict the 
consequences of alternative management practices, would extend its utility even more. 

This report reviews the various electrical conductivity methods for determining soil 
salinity, for monitoring it and for mapping it, along with methodology for establishing the 
locations of measurement sites. Advantages and limitations of the alternative methods are 
discussed, including their relative costs; practical integrated mobile-systems for 
measurement/monitoring/ mapping applications are also described. Examples of the utility of the 
various methods are given for mapping and monitoring soil salinity, for diagnosing saline seeps, 
for evaluating the adequacy and appropriateness of irrigation and drainage management, for 
scheduling and controlling irrigations, for determining the leaching needed to reclaim saline soils, 
and for locating areal sources of over-irrigation and salt-loading. For earlier reports on 
instrumental field methods of soil salinity measurement and assessment see Rhoades (1976, 1978, 
1984 and 1990a, b, 1992a, 1993, 1996a), Rhoades and Corwin (1984, 1990b), Rhoades and 
Miyamoto, (1990), Rhoades and Oster (1986) and Corwin and Rhoades (1990). 



Soil salinity assessment 



Chapter 2 

Determination of soil salinity from aqueous 

electrical conductivity 



The term salinity refers to the presence of the major dissolved inorganic solutes (essentially Na + , 
Mg ++ , Ca ++ , K + , CI, S0 4 = , HC0 3 , N0 3 and C0 3 = ) in aqueous samples. As applied to soils, it 
refers to the soluble plus readily dissolvable salts in the soil or, operationally, in an aqueous 
extract of a soil sample. Salinity is quantified in terms of the total concentration of such soluble 
salts, or more practically, in terms of the electrical conductivity of the solution, because the two 
are closely related (US Salinity Laboratory Staff, 1954). 



Principles of Aqueous Electrical Conductivity 

Electrical conductivity (EC) is a numerical expression of the inherent ability of a medium to carry 
an electric current. Because the EC and total salt concentration of an aqueous solution are closely 
related, EC is commonly used as an expression of the total dissolved salt concentration of an 
aqueous sample, even though it is also affected by the temperature of the sample and by the 
mobility, valences and relative concentrations of the individual ions comprising the solution 
(water itself is a very poor conductor of electricity). Furthermore, not all dissolved solutes exist 
as charged-species; some are non-ionic and some of the ions combine to form ion-pairs which are 
less charged (they may even be neutral) and, thus, contribute proportionately less to electrical 
conduction than when fully dissociated. 

The determination of EC generally involves the physical measurement of the materials' 
electrical resistance (R), which is expressed in ohms. The resistance of a conducting material 
(such as a saline solution) is inversely proportional to its cross-sectional area (A) and directly 
proportional to its length (L). Therefore, the magnitude of the measured electrical resistance 
depends on the dimensions of the conductivity cell used to contain the sample and of the 
electrodes. Specific resistance (R s ) is the resistance of a cube of the sample 1 cm on edge. 
Practical cells are not of this dimension and measure only a given fraction of the specific 
resistance; this fraction is the cell constant (K = R/R s ). 

The reciprocal of resistance is conductance (C). It is expressed in reciprocal ohms, i.e., 
mhos. When the cell constant is applied, the measured conductance is converted to specific 
conductance (i.e., the reciprocal of the specific resistance) at the temperature of measurement. 
Often, and herein, specific conductance is referred to as electrical conductivity, EC: 

EC = 1/R = K/R. [1] 



Determination of soil salinity from aqueous electrical conductivity 



Electrical conductivity has been customarily reported in micro-mhos per centimetre 
(umho/cm), or in milli-mhos per centimetre (mmho/cm). In the International System of Units (SI), 
the reciprocal of the ohm is the siemen (S) and, in this system, electrical conductivity is reported 
as Siemens per metre (S/m), or as decisiemens per metre (dS/m). One dS/m is equivalent to one 
mmho/cm. 

Electrolytic conductivity (unlike metallic conductivity) increases at a rate of approximately 
1.9% per degree centigrade increase in temperature. Therefore, EC needs to be expressed at a 
reference temperature for purposes of comparison and accurate salinity expression; 25° C is most 
commonly used in this regard. The best way to correct for the temperature effect on conductivity 
is to maintain the temperature of the sample and cell at 25° ± 0.5 °C while EC is being measured. 
The next best way is to make multiple determinations of sample EC at various temperatures both 
above and below 25° C, then to plot these readings and interpolate the EC at 25° C from the 
smoothed curve drawn through the data-pairs. For practical purposes of agricultural salinity 
appraisal, EC is measured at one known temperature other than 25° C and then adjusted to this 
latter reference using an appropriate temperature-coefficient (f t ). These coefficients vary for 
different salt solutions but are usually based on sodium chloride solutions, since their temperature 
coefficients closely approximate those of most salt-affected surface, ground, and soil waters. 
Another limitation in the use of temperature coefficients to adjust EC readings to 25° C is that 
they vary somewhat with solute concentration. The lower the concentration, the higher the 
coefficient, due to the effect that temperature has upon the dissociation of water. However, for 
practical needs, this latter limitation may be ignored and the value of/, may be assumed to be 
single-valued. It may be estimated as: 

ft = 1 - 0.20346 (T) + 0.03822 (T 2 ) - 0.00555 (T 3 ), [2] 

where T = [temperature in degrees Celsius - 25] /10. This relation was derived from data given in 
Table 15 of Handbook 60 (US Salinity Laboratory Staff, 1954). In turn, the electrical 
conductivity at 25° C, EC 25 , is calculated as: 

EC 25 =/*EC t , [3] 

where EC, is the EC at the measured temperature t. 

The above approach and/ - temperature relations have been routinely used to reference 
soil electrical conductivity values (Rhoades, 1976), as well as solution/extract conductivities. The 
applicability of these / ( factors were tested for their appropriateness in this regard and concluded 
to be appropriate by McKenzie, et al. (1989), Johnston (1994), and Heimovaara (1995). 

Because of differences in the equivalent weights, equivalent conductivities, and variations 
in the proportions of the various solutes found in soil extracts and water samples, the 
relationships between EC and total solute concentration and osmotic potential are only 
approximate. However, they are still quite useful. These relationships are as follows: total cation 
(or anion) concentration, mmoles charge/litre =10 x EC 25 , in dS/m; total dissolved solids, mg/litre 
= 640 x EC 25 , in dS/m; and osmotic potential, M Pa at 25° C = 0.04 x EC 25 , in dS/m. 



Soil Water Salinity 

Theoretically, the electrical conductivity of the soil solution (EC„ ) is a better index of soil salinity 
than is the traditional index (EC e ). This is so because the plant roots actually experience the soil 



Soil salinity assessment 



solution; they extract their nutrients from it, absorb other solutes from it and they consume this 
water through the process of transpiration. However, EC W has not been widely used as a means 
for measuring or expressing soil salinity for several reasons. Firstly, it is not single-valued; it 
varies over the irrigation cycle as the soil water content changes (Rhoades, 1978). Thus, EC W 
does not lend itself to simple classifications or standards unless it is referenced to a specific water 
content, such as field capacity. Secondly, and probably most importantly, EC„ has not been 
widely adopted for routine appraisals of soil salinity because methods for obtaining soil water 
samples are not practical at typical field water contents. 

Samples of soil solutions may be obtained from soil samples in the laboratory by means of 
displacement, compaction, centrifugation, molecular adsorption and vacuum- or pressure- 
extraction methods. The latter methods are described by Richards (1941); displacement methods 
by Adams (1974); combination displacement/centrifugation methods by Gillman (1976), 
Mubarak and Olsen (1976, 1977) and Elkhatib et al. (1986); a combination 
vacuum/displacement method by Wolt and Graved (1986); a simple field-pressure filtration 
method by Ross and Bartlett (1990); adsorption techniques by Davies and Davies (1963), 
Yamasaki and Kishita (1972), Gillman (1976), Dao and Lavy (1978), Kinniburgh and Miles 
(1983) and Elkhatib et al. (1987). Comparisons of the various methods have been made by 
Adams et al. (1980); Kittrick (1983); Wolt and Graved (1986); Menzies and Bell (1988) and 
Ross and Bartlett (1990). 

Two means of measuring EC„ in undisturbed soils exist. One is collect a sample of soil 
water using an in-situ extractor and then to measure its EC; the second is to measure EC„ 
"directly" in the soil using in-situ, imbibition - type "salinity sensors". 

A typical vacuum extractor system used to collect soil water samples in the field is shown 
in Figure 1. This suction-method, first proposed by Briggs and McCall (1904), is useful for 
extracting water from the soil when the soil-water suction is less than about 0. 1 M Pa. Although 
the available range of soil moisture for crops extends to 1.5 M Pa of soil suction, most water 
uptake by plants takes place within the zero to 0.1 M Pa range. Therefore, the suction method is 
applicable for many salinity-monitoring needs. While different extraction devices have been used, 
the most commonly used is the porous ceramic cup. Early vintage extractor construction and 
performance are described in a bibliography assembled by Kohnke et al. (1940). Reeve and 
Doering (1965) described the more modern equipment and procedures for its use in detail. These 
procedures have been used in field experiments with good success for salinity appraisal purposes. 
Wagner (1965) used similar devices to estimate nitrate losses in soil percolate. Other, improved 
and specialized, versions have since been developed for various purposes, including the following: 
a miniature sampler which eliminates sample transfer in the field (Harris and Hansen, 1975), 
samplers which shut off automatically when the desired volume of sample is collected (Chow, 
1977), samplers which function at depths greater than the suction lift of water (Parizek and Lane, 
1970; Wood, 1973) and samplers which minimize "degassing" effects on solution composition 
(Suarez, 1986, 1987). Soil water has also been extracted using cellulose-acetate hollow fibres 
(Jackson et al. 1976; Levin and Jackson, 1977), which are thin-walled, semipermeable, and 
flexible. Claimed advantages include flexibility, small diameter, minimal chemical interaction of 
solutes with the tube matrix, and compositional results comparable with those from samples 
obtained from ceramic extraction cups. Collection "pan"-type collectors have also been used to 
collect soil percolate (Jordan, 1968). Additionally, large-scale vacuum extractors (15 cm wide by 
3.29 m long) have been built and used to assess deep percolation losses and chemical composition 
of soil water (Duke and Haise, 1973). Ceramic "points", which absorb water upon insertion into 
the soil, have also been used to sample soil water with some success (Shimshi, 1966). However, 



Determination of soil salinity from aqueous electrical conductivity 



FIGURE 1 

Diagram of vacuum extractor apparatus for sampling soil water (after Rhoades and Oster, 

1986) 



VACUUM GAUGE 




V4CUU-M 



NEDPRCHE TUBING 



VACUUM 
TANK 




MANIFOLD SAMPLE 90TTLE5 



rmj.UH: c\)i 



SAMPLING TUBES 



only very small samples are obtained with these "points" and there are potential errors due to 
vapor transfer and chromatographic separation. Tadros and McGarity (1976) have analogously 
used an absorbent sponge material. 



Various errors in sampling soil water can occur with the use of any of the above types of 
extractors. Included are factors related to sorption, leaching, diffusion, and sieving by the cup 
wall; also to sampler intake rate, plugging, and sampler size. Nielsen et al. (1973), Biggar and 
Nielsen (1976), and van De Pol et al. (1977), used soil water extractors to determine salt flux in 
fields and have demonstrated that field variability in this regard is very large. They concluded that 
soil water samples being "point samples" can provide only indications of relative changes in the 
amount of solute flux, but not of quantitative amounts, unless the frequency distribution of such 
measurements is established. Because the composition and concentration of soil water is not 
homogeneous through its entire mass; water drained from large pores at low suctions (such as 
that collected by vacuum extractors) may have compositions very different from that extracted 
from micropores. A point source of suction, such as a porous cup, samples a sphere of different- 
sized pores dependent upon distance from the point, the amount of applied suction, the hydraulic 
conductivity of the medium, and the soil water content. Although vacuum extractors are versatile, 
easily usable and provide for in situ sampling of soil water, they have, as evident from the above 
discussion, limitations. For more discussion of the different suction-type samplers and other 
methods for sampling soil solution and various errors associated with them see the reviews by 
Rhoades (1978, 1979a), Rhoades and Oster (1986), litaor (1988) and Grossman and Udluft 
(1991). 



Soil salinity assessment 



FIGURE 2 

(A) Imbibition-type salinity sensor with spring, housing and pin in disassembly; (B) schematic 

of internal elements of salinity sensor (after Rhoades and Oster, 1986) 




ELECTRO. fT*: 
Fl EUENT-* 



PLATINUM 
ELECTRODES 



THEPAKlSTOP 




B 



INNER ASSEMBLY HOUSING 



When the total concentration of salts in the soil water is sufficient information, i.e., when 
specific solute analyses are not needed, in-situ devices capable of directly measuring EC W may be 
used advantageously. Kemper (1959) developed the first in situ salinity sensor. It consisted of 
electrodes imbedded in porous ceramic to measure the electrical conductivity (EC) of the solution 
within the "ceramic cell". When placed in soil, these devices imbibe water which, in time, comes 
to diffusional equilibrium with the soil water. Richards (1966) improved the design of the soil 
salinity sensor to shorten its response time and to eliminate external electrical current paths. This 
unit is now produced commercially. In this unit (Figure 2), the salinity sensitive element is an 
approximately 1 -mm- thick ceramic plate which contains platinum screen electrodes on opposite 
sides. This gives a short diffusion path and thus lowers response time. Another feature of the 
design is a preloaded spring. After the salinity sensor is placed in the soil, the spring is released to 
ensure good contact of the ceramic plate with the soil. A thermistor is incorporated in the sensor 
so that the EC may be adjusted for temperature effects. A commercially available meter 
developed for these sensors is shown in Figure 3. An oscillator circuit system has been developed 



10 Determination of soil salinity from aqueous electrical conductivity 



FIGURE 3 

Commercial meter and salinity sensor showing ceramic disc in which platinum electrodes are 

embedded (after Rhoades, 1993) 




for automated salinity sensor measurements and data logging (Austin and Oster, 1973). This 
permits linear readings to be obtained with lead lengths of up to several hundred meters. 

Salinity sensors have been mostly used in agricultural research, where continuous 
monitoring of soil salinity in soil columns, lysimeters, and field experiments is required (Oster 
and Ingvalson, 1967; Rhoades, 1972; Oster et al. 1973; Oster et al. 1976; Ingvalson et al. 1970). 
The accuracy of the commercial ceramic sensor under such conditions has been found to be ± 0.5 
dS/m (Oster and Ingvalson, 1967). Reliability was determined by removing the sensors from field 
and lysimeter experiments after 3-to-5 -years of continuous operation and comparing their 
calibrations relative to original ones (Oster and Willardson, 1971; Wood, 1978). About 68 
percent of the tested sensors had calibrations within 14 percent of the original calibrations after 
five years. Shifts varied in direction and magnitude, and some complete failures occurred. 

Response times of the commercial salinity sensors have been evaluated in field situations 
(Wesseling and Oster, 1973; Wood, 1978). In the matric potential range of -0.05 to -0.15 M Pa, 
90 percent of the response of these sensors to a step change in salinity will occur within 2 to 5 
days. At lower matric potentials, response times are longer. Thus, it may be concluded that 
salinity sensors are not well suited for measuring short-term changes in salinity because of their 
relatively long response time of at least several days. Desaturation of the ceramic occurs at matric 
potentials more negative than -0.2 M Pa, significantly reducing the conductance of the ceramic 
salinity sensor (Ingvalson et al. 1970). Hence, this type of sensor is not accurate in "dry" soils. 
Salinity sensors constructed of porous glass have been developed which remain saturated with 
soil water to 2 M Pa matric potentials (Enfield and Evans, 1969), but they are fragile and not 
available commercially. 



Soil salinity assessment 



11 



FIGURE 4 

Variations in soil water electrical conductivity (EC W ) and soil water tension in the root zone of 

an alfalfa crop during the spring of the year (after Rhoades, 1972) 



16 - 



T 




o 

X 
E 



9 19 
Fab 



II 21 91 
March 



10 20 30 10 20 30 9 19 29 

April May June 






Soil disturbance during installation can result in errors associated with modified water 
infiltration in the back-filled hole used to install salinity sensors. Special precautions during their 
installation must be taken to avoid this. 

While, obviously, also with limitations, salinity sensors may be used advantageously for 
continuously monitoring electrical conductivity of soil water at selected depths over relatively 
long periods of time, as illustrated in Figure 4. They are not well suited for measuring short-term 
changes of salinity, especially in "dry" soils. Many units may be needed because of their small 
sampling volume, and the substantial heterogeneity of soils, in order to characterize the actual 
conditions existing in irrigated soils. These numbers can be minimized if the sensors are primarily 
used to follow changing salinity status at a specific location over time. They are simple in 
principle, easily read, and sufficiently accurate for intermediate-term salinity monitoring 
purposes. They must be individually calibrated; these calibrations may change with time. They 
are, of course, not practical for mapping purposes for obvious reasons. 



Soil Extract Salinity 



Because present methods of obtaining soil water samples at typical field water contents are not 
very practical, aqueous extracts of the soil samples have traditionally been made in the laboratory 
at higher- than-normal water contents for routine soil salinity diagnosis and characterization 
purposes. Since the absolute and relative amounts of the various solutes are influenced by the 
water/soil ratio at which the extract is made (Reitemeier, 1946), the water/soil ratio used to 
obtain the extract should be standardized to obtain results that can be applied and interpreted 
reasonably generally. As stated earlier, soil salinity is most generally defined and measured on 
aqueous extracts of so-called, saturated soil-pastes (US Salinity Laboratory Staff, 1954). The 
water content of saturated soil-pastes (the so-called saturation percentage, SP), as well as the 



12 



Determination of soil salinity from aqueous electrical conductivity 



FIGURE 5 

(A) Saturated soil-paste in mixing container and 
in drying/weighing container; (B) saturated soil- 
paste being extracted by vacuum; and (C) 
collection of saturated-paste extract 





water/soil ratio, varies with soil texture. 
It is related in a reasonably general and 
predictable way to soil- water contents 
under field conditions. For these same 
reasons, crop tolerance to salinity is also 
most generally expressed in terms of the 
electrical conductivity of the saturation- 
extract (EC e , Maas and Hoffman, 1977; 
Maas, 1986, 1990). Herein, the term 
saturated soil-paste extract is often used 
in place of saturation-extract; the two 
terms are synonymous. 

Estimates made of the EC W from 
EC e and the ratio of their water contents 
will usually be excessively high. This is 
because salts will often be present in the 
saturation-extract that would not be 
under actual field conditions. Addi- 
tionally, salts contained within the fine 
pores of aggregates will contribute to the 
EC e value, though it is doubtful that 
significant amounts of such salts are 
absorbed by plant roots or affect the 
availability of the majority of the water 
extracted by the plant (which is 
primarily that present in the larger 
pores). 

EC e is typically determined as 
follows. A saturated soil-paste is 
prepared by adding distilled water to a 
sample of air dry soil (200 to 400 g) 
while stirring and then allowing the 
mixture to stand for at least several 
hours (but often overnight) to permit the 
soil to fully imbibe the water and the 
readily soluble salts to fully dissolve, so 
as to achieve a uniformly saturated and 
equilibrated soil-water paste (see Figure 
5 A). At this latter point, which is 

sufficiently reproducible for practical purposes, the soil paste glistens as it reflects light, flows 
slightly when the container is tipped, slides freely and cleanly off a spatula, and consolidates 
easily when the container is tapped or jarred after a trench is formed in the paste with the broad 
side of the spatula. The extract of this saturation-paste is usually obtained by suction using a 
funnel and filter paper (see Figure 5B and 5C). The EC and temperature of this extract are then 
measured using standard conductance meters/cells and thermometers, respectively (see Figure 6); 
the EC] 5 value of this extract is calculated from Equation [3] to give EC e . 




3^^^^^^^^» .- ^^^ 



Soil salinity assessment 



13 



Once soil extract samples are 
obtained, laboratory chemical analyses 
can be carried out to determine, in 
addition to the electrical conductivity of 
the extract (EC e ), the concentrations of 
the individual solutes, i.e., Na + , Ca ++ , 
Mg ++ , K\ C1-, S0 4 \ HCO3-, C0 3 = , 
NO3-, etc. Methods for such analyses 
are given elsewhere (Rhoades, 1982; 
Soil Science Society of America, 1996). 
More details about the methods for 
measuring the electrical conductivity and 
total dissolved solids contents of 
aqueous samples and extracts are given 
in Rhoades (1982, 1993, 1996a); for a 
good discussion of some of the 
operational factors influencing the 
procedure see Shaw (1994). 

Though the above-described 
procedure for making a saturated soil- 
paste is somewhat subjective, diagnoses 
are not compromised by the normal 
variations experienced in it. Yet this 
subjectivity seems to be a concern to 
some people (Shaw, 1994). To eliminate 
some of the subjectivity of the saturation 
extract method, Longenecker and Lylerly 
(1964) proposed wetting the sample by 

capillarity using a "saturation table". Beatty and Loveday (1974) and Loveday (1972) advocated 
predetermining the amount of water at saturation on a separate soil sample using a similar 
capillary wetting technique and then adding this amount to all other samples of the same soil. 
Allison (1973) recommended slowly adding soil to water, rather than water to soil, when making 
pastes to speed wetting of the soil and preparation of the saturated-paste condition. All of these 
modifications offer advantages over the standard procedure under certain situations, but all but 
the last one slow the procedure considerably without significantly enhancing the diagnostic value 
of the result. 




Other extraction ratios, such as 1:1, 1:5, etc., are easier to use than that of the saturation 
paste but they are less well-related to meaningful soil properties and are more subject to errors 
resulting from peptization, hydrolysis, cation exchange, and mineral dissolution. Sonnevelt and 
van den Ende (1971) recommended a 1:2 volume extract. This method is a compromise between 
the saturation-paste extract and the higher-dilution "weight" extracts. The water contents of the 
1:2 volume "pastes" of sandy and clayey soils are higher and lower, respectively, relative to the 
saturation-paste extract. For purposes of monitoring, when relative changes are of more concern 
than the absolute solute concentration(s), these quicker, simpler methods of "fixed-extraction- 
ratios" may be used to advantage in place of the saturation extract. Of course, the relations given 
in Handbook 60 to predict exchangeable sodium percentage from the sodium adsorption ratio 
apply only to the saturation-paste extract, as do most of the other indices/criteria/ standards used 
to express/interpret soil salinity/sodicity/toxicity and plant response (salt-tolerance, plant-growth 



14 Determination of soil salinity from aqueous electrical conductivity 



data) from soil analyses. Criteria for evaluating salinity and sodicity effects on soils and crops 
have been developed for some of the higher-dilution extracts, but the greater errors and lack of 
uniformity they create in this regard makes the extrapolation of results more difficult and the 
literature more confusing (Rengasamy, 1997). Consistency and uniformity of methodology and 
criteria/standards should be sought whenever possible to facilitate interpretation of results and 
their general applicability. 

Because of the numerous interacting effects of the following: mineral dissolution/ 
precipitation, cation exchange, ion-pair formation, negative adsorption, time of equilibration, 
amount of grinding and drying, presence or absence of suspended minerals and organic matter in 
the extract, microbial production of C0 2 during equilibration, etc., a computer deterministic- 
chemical model is required, along with the determination of the ionic-composition of the extract, 
of the associated cation exchange composition, of the cation-exchange-capacity and the cation 
exchange coefficients, in order to accurately calculate the EC of a solution in association with soil 
as the water/soil ratio is altered (Paul et al., 1966). The assumption of conservation of mass with 
change of water content during a change in water content is not sufficiently valid to permit the 
EC at a second water content to be accurately calculated as the product of the EC at the second 
water content and the ratio of the two water contents. Since the computer-model approach is too 
demanding and the second ratio approach too simplistic, various empirical relations have been 
developed to estimate EC e from EC values measured on higher-dilution extracts. Shaw (1994) 
has reviewed most of these methods and concluded that they are very location-specific and can 
not be extrapolated reliably elsewhere. He developed an improved, more generally applicable 
relation to estimate EC e from the EC of a 1:5, soil: water extract, but it requires an analysis of the 
extract for chloride concentration and the determination of the air dry moisture content of the soil 
sample used to make the 1:5 extract. This method and all of the others like it will be more 
accurate for solutions dominated by chloride salts; very substantial errors may occur with soils 
containing gypsum, especially if they are also sodic (Adiku et al., 1992). One must question 
whether any of these "conversion" methods save sufficient time and effort to make them 
worthwhile, especially considering the uncertainty in their resulting estimates of EC e . A faster and 
more accurate method for estimating EC e , based on simple measurements of the volume- weight 
and EC of the saturated-paste itself (rather than of its extract), is described in the following 
chapter. This method eliminates much of the work involved in measuring EC e and SP using 
conventional methods (the latter involves oven-drying for 24 hours) without the loss of accuracy 
that occurs in estimating it from EC measurements made on extracts obtained at higher dilutions. 
Additionally, one obtains the added SP information with this method, which is valuable as an 
estimator of many soil properties including texture, water-holding capacity and cation-exchange 
capacity. 



Soil salinity assessment 1 5 



Chapter 3 

Determination of soil salinity from soil-paste 
and bulk soil electrical conductivity 



The methods of soil salinity determination already described are not well suited for use in the field 
nor for intensive-mapping and monitoring applications because they require the collection of soil 
samples and their aqueous extracts. Thus, they are relatively slow and expensive to carry out (see 
later discussions and Chapter 5). For this reason, more practical field methods have been sought 
and developed. One such method eliminates the need for aqueous extractions, though it still 
requires the collection of soil samples and the making of saturated soil-pastes. Another still more 
practical method is based on direct measurements of bulk soil electrical conductivity (EC a ) made 
upon undisturbed soils using geophysical- type sensors; this methodology is especially well suited 
for intensive mapping and monitoring applications. These two methods are described in this 
section. 



Principles of Soil and Soil-Paste Electrical Conductivities 

A model of the electrical conductivity of mixed soil/water systems that has been shown to be very 
useful and generally applicable for purposes of salinity appraisal is illustrated in Figure 7. This 
model supersedes the earlier model of Rhoades et al. (1976). It assumes that the electrical 
conductivity of a soil containing dissolved electrolytes (salts) in the soil "solution" can be 
represented by conductance via the following three pathways (or elements) acting in parallel: (1) 
conductance through alternating layers of soil particles and the soil solution that envelopes and 
separates these particles (a solid-liquid, series-coupled element), (2) conductance through 
continuous soil solution pathways (a liquid element), and (3) conductance through or along the 
surfaces of soil particles in direct and continuous contact with one another (a solid element). 

Because most soil minerals are insulators, electrical conduction in sufficiently moist soils is 
primarily via the electrolytes (salts) contained in the water occupying the larger pores. The 
contribution of the solid phase to electrical conduction in moist soils, the so-called surface 
conduction, is primarily via the exchangeable cations associated with the clay minerals (though 
the latter are actually present in the aqueous phase). Surface conductance is generally smaller 
than that of the pore-solution because the former electrolytes are more limited in their amounts 
and mobilities. The magnitude of surface conduction is assumed (and tests have confirmed this) 
in the soil electrical conductivity model to be, for practical purposes, independent of the dissolved 
salts and essentially constant for any given soil (Rhoades et al., 1976; Shainberg et al., 1980; 
Bottraud and Rhoades, 1985a). The surface conductance is also assumed to be coupled in series 
with the electrolyte present in the water films associated with the solid surfaces and in the small 



16 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



FIGURE 7 

Model and schematic representation of electrical conduction in soil: (A) the three paths that 

current can take in unsaturated soil; (B) simplified soil electrical conductance model 

consisting of the three conductance elements (a-c) acting in parallel (after Rhoades ef a/., 

1989a) 



WW Hi 




C.T 



water-filled pores which serve as "links" between adjacent particles and aggregates to provide a 
secondary pathway for current flow in moist soils. This pathway is modelled as acting in parallel 
with the primary, continuous pathway (salt-solution contained in the large water-filled pores). 
The solid element pathway may exist in soils with indurated layers. In such layers, conductance 
could occur through, or along the surfaces of, the soil particles, which are in direct and 
continuous contact with one another (a solid element). The relative flow of current in the three 
pathways depends upon the volumetric contents and solute concentrations of the water in the two 
different categories of pores and on the volumetric contents and magnitudes of the surface- 
conduction and of the indurated solid-phase. This model is mathematically represented by 
Equation [4]: 



EC„ = 



(e ss + e ws ) 2 ec ks ec ss 

(Q SS )EC KS +(Q KS )EC S 



+ 0„, EC,, + ©..„ EC, 



[4] 



where EC a is the electrical conductivity of the bulk soil; @ ws and @ wc are the volumetric soil 
water contents in the series-coupled pathway (small pores) and in the continuous liquid pathway 
(large pores), respectively; @ ss and @ sc are the volumetric contents of the surface-conductance and 
indurated solid phases of the soil, respectively; EC„ S and EC WC are the specific electrical 
conductivities of the soil water that are in series-coupling with the solid particles and in the 
continuous conductance element, respectively, and EC SS and EC SC are the electrical conductivities 
of the surface-conductance and indurated solid phases, respectively. The soil water in the 
continuous pathway is envisioned as the water occupying the larger pores, commonly referred to 
as "mobile" water. This water can be different in composition from that in the small pores and 



Soil salinity assessment 1 7 



intra-ped pores, which is envisioned as the "immobile" water associated in the model with the 
series-coupled pathway. Ultimately, diffusion processes will cause EC WS and EC„ C to be equal. 
However, when water is being added by irrigation or rain, or is being removed by drainage or 
evapotranspiration, equilibrium will not exist; consequently, EC WS and EC WC may be different 
during these periods. 

The second term of the second member in Equation [4], i.e. @ sc EC SC , usually may be 
dropped. This is so, apparently, because soil structure simply does not allow for enough direct 
particle-to-particle contact between aggregate units in typical agricultural soils to provide a 
continuous solid-phase pathway for electrical current flow. This latter potential, pathway is 
disrupted by water films surrounding the particles and peds or by void spaces within the matrix 
that are filled with either liquid or air. Experimental data show it to be negligible (Rhoades et al., 
1976, 1990a). Thus for all but soils with indurated layers, Equation [4] may be simplified to the 
following two-pathway model (Rhoades et al. 1989a): 



EC = 



(Q S +® WS ) 2 EC WS EC S 
(® S )EC WS +(® WS )EC S 



+ (& W -Q WS )EC WC , [5] 



where (©„ - @ ws ) is substituted for @ wc , W is the total volumetric soil water content, and EC S is 
the surface conductance of soils without indurated layers. This equation has been shown to be 
generally applicable to arid-land mineral soils of the Southwestern United States (Rhoades et al., 
1989a, 1990a). There is no reason to believe that it is not equally applicable to similar arid-land 
soils found elsewhere in the world. However, the model has not been tested on soils containing 
high contents of gypsum, which may differ because gypsum particles may be more conductive 
than silicate mineral particles. This could result in higher values of EC SS and EC SC . This problem 
has not been observed by the author in gypsiferous US soils, but they do not contain as high of 
gypsum contents as occur in some parts of the Near East; no reports have been found indicating 
that others have observed this to be a significant problem. 

For conditions of EC„ , greater than about 2-4 dS/m and for soils with typical values of EC S 
(less than about 1.5 dS/m), the product (@ s * EC WS ) is so much larger than the product (@ ws * 
EC S ) that the latter product can be neglected; thus simplifying Equation [5] to the following one 
for typical saline soils: 



EC = 



(e s +@ ws y ec s 



+ (& W -& WS )EC WC . [6] 



Equation [5] is the more generally applicable relation and must be used for soils with low 
values of EC WS , i.e. non-saline soils, where the relation between EC a and EC„ C is curvilinear at 
low levels of EC„ c . The first term of the second member of this equation determines the shape of 
the nonlinear portion of the EC a -EC„ c curve. Over the remainder of the EC WC range, EC a and EC WC 
are linearly related, with (@ w - ® ws ) representing the slope of this relation. In contrast, the 
simplified relation expressed in Equation [6] should only be used for conditions of EC„ c > about 2 
dS/m (EC e > about 1-2 dS/m) and EC S < 1.5 dS/m, i.e., for typical saline soils. For such cases, 
the relation between EC a and EC„ c expressed in Equation [6] is linear and proportional to (©„ - 
@ ws ) beyond the threshold value of EC WC and the y-intercept depends upon EC S , @ s and @ ws . Since 
the ratio [(@ s + ws ) 2 / S ] is typically close to the value 1 (because @ s is typically about 0.5 and 
@ ws is less than or equal to 0.5 @„ , where @ w is typically about 0.4 or less); the intercept of 



18 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 




WgjkeitQ loam 

Sj ■ 0.634 
EC, » 0.43O 



Equation [6] is approximately equal to 
EC S and may be symbolized as EC S *. 
The earlier EC a model of Rhoades, et al. 
(1976) is analogous to this limiting case 
version of Equation [6], as shown 
elsewhere (Rhoades et al. 1989a). This 
earlier model expressed the slope in 
terms of a tortuosity concept, but it is 
mathematically identical to that 
expressed in Equations [5] and [6] which 
supersede it. The major improvement is 
contained within the intercept term of 
Equation [5]. 

Data illustrating the appropriate- 
ness of the above described model and 
generalizations are shown in Figure 8 for 
Waukena loam soil. The solid line is that 
described by Equation [5], the dashed 
line is that described by Equation [6] for 
the one example water content (0.375, 
for purposes of illustration), and the 
circles represent experimental data. EC„ 
represents the EC of the equilibrating 
water or the water extracted from the 
soil by pressure filtration. Note that 
salts, as well as water, were removed 

during the pressure filtration of the soil that was used in this controlled-experiment to vary water 
content while keeping EC„ constant. The soil had been extensively leached with waters of 
different salinities (EC„ values), therefore EC WC and EC WS were essentially equal to EC W under the 
conditions of this experiment. These data and the model relations also show that EC„ can be 
inferred from measurements of EC a made at relatively low water content, but the ability to 
accurately do so decreases as @ w decreases. This is so because the required accuracy of 
measurement of EC a becomes limiting as the EC a = f (EC„ ) relation flattens at low values of W 
(Rhoades et al., 1976; Bottraud and Rhoades, 1985b). However, at very low values of @ w , it is 
not possible to determine EC W (or EC e ) from EC a at all. The value of the "threshold" water 
content is approximately 0.10. For more discussion and data about the threshold water content 
see Figure 6 and Table 3 in Rhoades et al. (1976). "Dry" soil measurements are to be avoided for 
the reason given above and for those which follow. 



FIGURE 8 

Electrical conductivity of Waukena loam soil as a 
function of the electrical conductivity and 
volumetric content of soil water. The measured 
data points (o) are shown and the solid line is the 
"fit" of these combined data by Equation [5] (after 
Rhoades et al., 1989a) 



to 

■o 

u 

o 
> 






Ld 



°0 4 B 12 IB zo 

Electrical Conductivity of Sail Waler, EC W] dS/m 




As explained above, the ability to accurately determine EC„ (or EC e ) from EC a decreases 
as @ w decreases. For this reason, it is recommended that EC a measurements be limited to 
moisture contents that are not less than about one-half of field-capacity water content. Most 
irrigated soils are kept above this level during the cropping season. As also stated above and 
elsewhere (Rhoades et al., 1976), it is not possible to measure EC a at very low values of @ w ; nor 
is it possible to use measurements of EC a to determine salinity under such conditions. This is so 
because there must be a continuous pathway for electrical flow through the soil in order to make 
the measurement. It should be noted that the soil-EC model assumes the presence of sufficient 
moisture to permit current flow to take place via the two pathways existing within the soil matrix 



Soil salinity assessment 



19 



(the water phase which is in continuous contact via the larger soil pores and the water films 
which envelope and bridge soil particles to form another continuous pathway). As explained 
above, the threshold value of @ w required to satisfy the above requirement is about 0.1, possibly 
more in sandy soils. This minimum limit will usually be met in all but the surface dry-mulch layer 
of irrigated soils during most of the irrigation season. However, it is another matter for dryland 
soils. Since dry soil is essentially an insulator, no useful information about salinity, or other soil 
properties for that matter, can be inferred from EC a measurements made on such dry soils. 
Therefore, one should not include the dry surface mulch in samples used to calibrate EC a - soil 
properties. EC a measurements should only be made in dryland soils during the time of the year 
when they are sufficiently moist for the measurable-conduction of electricity. It is sufficiently 
important to repeat: it is inappropriate to try to infer salinity from measurements of EC a 
made on dry, or nearly dry, soil as it is to include salinity analyses of such soils in the data 
used to establish EC, - EC e calibrations . This will be commented on later when the relative 
merits of the different sensors which can be used to measure EC a , as well as the different methods 
of calibration, are discussed. 



Since, @ s = p b / p s , soil bulk density (p b ) and soil particle density (p s ) are two soil 
properties, besides salinity, that affect EC a . The value of EC a is also affected by clay content and 
type, since EC S is primarily associated with the cation exchange capacity. Additionally, EC a is 
expected to be affected by the pore size distribution and structure of the soil, since they influence 
the contents of "mobile" (@„ c ) and "immobile" (@ ws ) water. Likewise, prior events (i.e., 
irrigation, rainfall, evapotranspiration) and processes (i.e., diffusion) which influence the 
distributions of salt concentrations between the mobile and immobile phases (EC WC and EC„ S , 
respectively) can affect EC a . The sensitivity of EC a to each of these factors can be determined 
from Equation [5] to the extent that they can be related to the parameters used in the model, for 
example EC S = f (% clay, clay type), @ s and @ w = f (p b and p s ), etc. A sensitivity analysis of this 
equation was undertaken with emphasis on soil salinity appraisal (Rhoades et al., 1989c; 
Rhoades and Corwin, 1990). These findings are not reviewed herein. They show that the values 
of the soil parameters that can not be easily measured in the field (i.e., bulk density, particle 
density, clay percentage and total and 
"immobile" water contents) can be 
estimated sufficiently accurately for 
the purposes of practical soil salinity 
appraisal. 



To use Equation [5] or [6] to 
assess soil salinity (EC W or EC e ) from 
EC a , the values of EC S , @„ s and @ w 
must be known. EC S , @ w , and @ w c can 
be estimated using Figures 9 and 10, 
respectively. The means used to obtain 
these relations are described elsewhere 
(Rhoades et al. 1989a). @„ can be 
measured in the field using various 
methods, if salinity is not too high, or it 
can be adequately estimated, for our 
purpose, by an indirect method that 
will be described later. @ s can be 
estimated from bulk density ( p B ) as 



FIGURE 9 

Correlation between EC S and clay percentage for 
a number of soils from the San Joaquin Valley of 
California (after Rhoades et al., 1989a) 







ao m « so 

Clqy Content, c, % 



20 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



FIGURE 10 

Relationship between the volumetric 
content of soil water existing in the series- 
path (@ ws ), the continuous-path (0 W c), and 
the total water content (® w ) found for 
various California soils (after Rhoades ef a/., 
1989a) 



cs 



- as 



£ 






£ 
* 



(j 

,u 



F 



r 3 =0i9iJ45 



S = p b 12.65, where 2.65 is a reasonable 
estimate of the average particle density of 
most mineral soils. Bulk density can also be 
estimated sufficiently accurately for our 
purposes, as explained later. 

Another factor affecting EC a , which 
may be important in some situations, is soil 
temperature. The electrical conductivity of 
soils containing moisture increases 
approximately 2 percent per degree Celsius 
increase in temperature. To simplify the 
interpretation of soil salinity data, it is 
customary to determine the temperature at 
which the measurement of EC a is made, and 
then, by means of correction tables or 
equations, to convert the measurement to a 
reference temperature (Rhoades, 1976). The 
temperature factors (f t ) obtained from 
Equation [2] are suitable for this purpose 
(McKenzie et al, 1989; Johnston, 1994, and 
Heimovaara (1995). However, sometimes it is 
preferable to encompass the effect of 
temperature by including it within the calibration relation established to predict soil salinity for 
the particular field conditions that existed at the time the EC a measurements were made. 

Equation [5] may be solved for EC W , with the assumption that EC„ S = EC„ C , by arranging 
it in the form of a quadratic equation and solving for its positive root as: 




O.I O.Z 0.3 0.4 0.5 

Volumetric Content of Soil lAtaler. 0_ 



FC =- 



b+^b 2 -4ac 
2a 



[7] 



where a = [(0)(0 W - 0.,)], b = [(0, + WS ) 2 (EC S ) + (0. - WS ) (0 WS EC S ) - (0 EC,)], and 
c = [0 ws EC s ECJ. 

If EC e is desired, it can be estimated from: 



(EC WC WC + EC WS WS ) = EC. W = EC £ pb SP / 100, 



[8] 



where SP is the gravimetric water content of the saturated-paste expressed as a percentage, and 
p b is the bulk density of the soil (Rhoades, et al. 1989a). The latter derived relation is strictly 
valid only for chloride-salt systems. The errors inherent in this approximation are analogous to 
those discussed with reference to estimating EC e from the EC value of extracts obtained at higher 
dilutions. However, the errors involved in Equation [8] are smaller because of the lower water 
contents used to make the saturation extract (compared to higher water/soil extracts). Some data, 
which supports the approximate equality of Equation [8], is presented in Figure 11. It may well 
be that EC e values predicted from EC a are more appropriate estimations of soil salinity than 
conventionally measured values of EC e . This is so because the latter measurements are subject to 



Soil salinity assessment 



21 



the errors inherent to aqueous extracts previously discussed, because salts present within the 
"immobile" water contribute to EC e but not to the EC WC value, which the author considers the 
plant is more responsive to, and which mostly contributes to soil leachates. 

The close relation that exists between EC e and EC a (as observed by numerous investigators) is 
made more apparent by substituting the above-mentioned approximate identity (Equation [8]) into 
equation [5], which yields the following relation: 



EC = 



(0 +0 ) 2 EC EC 

\ s ws ) ws s 

EC +0,. EC 



+ 



©, 







V ~w j 



100 



EC. 



[9] 



The experimental data presented in Figures 
12 and 13 imply that this relation is 
generally applicable to arid-land soils. This 
relation also implies that the slope of EC, = 
f (EC e ) plots, and vice-versa, are related to 
soil type , since SP and p b vary with soil 
type (evidence of this is shown in Figures 
14 and 15). It also implies that the relation 
between EC a and EC e is not much 
influenced by variation in soil water 
content, as stated earlier, since the ratio @ w c 
/ @ l( is essentially a constant (~ 0.36; see 
Figure 10). The linear relations shown in 
Figures 13 and 14 were based on a 
relatively small data set of soils; they may 
be expected to be more curvilinear, like that 
shown in Figure 15, when a wider range of 
soil types are included. A curvilinear 
relation also has been reported by Johnston 
(1994). 

Equation [9] can be further simplified 
and approximated by substi-tuting into it 
the value 0.36 for @ l(C / ©„ found to be 
typical of California arid-land soils and the 
familiar approximation (U. S. Salinity 
Handbook 60, 1954), SP p b / 100 = 2 6 fc , 
where @ fc is the volumetric water content at 
field capacity, to give: 



FIGURE 11 

Relationship between the product of soil 
water electrical conductivity (EC W ) and 
volumetric water content (® w ) and the 
product of the electrical conductivity of the 
saturated-paste extract (EC e ), its saturation 
percentage (SP), and the soil bulk density 
(Pb ) found for soils of the Wellton-Mohawk 
Irrigation and Drainage District, Arizona, 
USA (after Rhoades, 1980) 




IQ 12 M « IB 



EC„ 



(0+0 Y EC EC 

V s ws } ws s 

EC, +0. EC 



+ 



(O.72)(0 /e )£C e 



[10] 



Evidence to support this relation is given in Figure 16. 



22 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



FIGURE 12 

Correlation between slopes of EC e vs. EC a 
calibrations obtained for different soils and their 
saturation percentage (SP) and bulk densities (p b ) 

(after Rhoades etal., 1989a) 



!D 



U 

Ld 






Q. 

Q 



<a 



<>. 




*S* ■ 0.03175* + 4.II5G* - 1.5021. 
wtwrt * = (I001/(SW B ) 
r 1 = 0.90 



JL 



2 3 

tioo)/{sp)yy 



j_ 



FIGURE 13 

Correlation between slopes of EC e vs. 
calibrations obtained for different soils 
saturation percentages (after Rhoades, 1981) 



EC a 
and 



The above two relations, along 
with Equation [6], provide the 
theoretical basis for the previously 
reported findings that the calibration 
relating EC a and EC e for any saline 
soil can be expressed as a simple 
linear equation of the following type: 

EC £ (or EC.) = m (EC a - EC S ')[11] 

for which the slope can be predicted 
from their field capacity water 
content, or their SP value, or their 
texture using relationships like those 
shown in Figures 12 to 16 and the 
intercept value (EC S , the intercept 
term of equations [5], [6], [9] and 
[10], essentially EC S ) can likewise be 
predicted from soil-texture related 
relationships such as that shown in 
Figure 9 (Rhoades, 1981; PJioades, 
et aL, 1989a, 1990a). Of course, 
empirical calibrations of the type 
expressed by Equation [11] may be 
obtained for soils using various direct 
methods. Simple field-procedures 
have been developed in order to 
obtain calibration relations 

appropriate to field soils with their 
particular natural structures, pore 
size distributions and water holding 
properties (Rhoades, 1976, 1980, 
1981; Rhoades and Ingvalson, 1971; 
Rhoades and van Schilfgaarde, 1976; 
Rhoades et cd. 1977). Two of these 
procedures are illustrated in Figures 
17 and 18 and are discussed in more 
detail in Annex 1 . 

A typical linear relationship 
between EC e and EC a of the type 
expected from Equations [9] to [11], 

as obtained by the direct calibration of an arid-land soil at near field-capacity water content, is 
shown in Figure 19. Analogous relations between EC W and EC a have been developed for some 
typical arid-land soils (see Figure 20, after Rhoades, 1980). With such calibrations, one can 
predict EC e (or EC„) from EC a for field soils of various types, provided they are in a sufficiently 
moist condition. While most of these calibrations have been developed for soils at or near field 
capacity water content at the time of EC a measurement, they have also been developed for soils 
under drier conditions and found not to differ substantially (Halvorson and Rhoades, 1974). 



ID 



9- 



6- 



K 4- 




Siope - - <HZZ06(5P)+ UiEEBa 



10 SO 30 40 



Soil salinity assessment 



23 



FIGURE 14 


Relationship found between saturation 


percentage and clay percentage for some 


California soils (after Rhoades et al., 1990a) 






ii ii i i t i i i / i 

o / 






10 


£P>1.±I9I1&CI • li.HO / 






ft ™ 


('•O.M / 






c 

41 

^ SO 

B 

£ 


/4 






| 4 ° 








1 * 

01 


/o 






io 


J 






ID 










< 


■ i j i > 






J 10 20 SO *0 50 60 70 


Clay content, C, % 





Numerous satisfactory field 
calibrations (like Figure 19) have been 
obtained for many soils around the world 
and they have been found to be similar for 
soils of similar textures (Rhoades and 
Ingvalson, 1971; Halvorson and Rhoades, 
1974; Rhoades, 1976, 1979a, 1980, 1981; 
Halvorson et al. 1977; Rhoades et al. 
1977, 1989a; Yadav et al. 1979; Loveday, 
1980; van Hoorn, 1980; Nadler, 1981; 
Bonn et al. 1982; Johnston, 1994). These 
calibrations have been found to be 
essentially independent of soil sodicity, 
provided soil structure and porosity have 
not been seriously degraded by the 
sodicity. Evidence of this is given in 
Figure 21 obtained in the controlled 
laboratory experiments of Bottraud and 
Rhoades (1985a). The sodium adsorption 
ratio (SAR) is a good estimator of soil 
sodicity (US Salinity Laboratory, 1954). 
Additional supportive evidence, though 
less rigorous and exact, is found in 
Shainberg et al. (1980) and Johnston 
(1994). 

An alternative model procedure for 
determining salinity from EC a at various 
water contents has been suggested by 
Nadler (1982). This procedure "curve-fits" 
what amounts to a f = (@) relation using 
moisture- tension data established for the 
particular soil in question and an empirical 
"effective porosity" relation based on A@. 
To date, the method has been successfully 
applied only to disturbed soil samples; it 
requires considerable laboratory effort to 
establish the empirical fit; it only applies 
to the "fitted" soil, and its applicability to 
field soils was found to be not generally 
good (unpublished data). 

The advocated model and 
experimental data (Rhoades et al., 1976, 
1989a, 1989c; Rhoades, 1990b) show that 

EC a is primarily a measure of the content of dissolved electrolyte present in a unit- volume of 
soil; note that the product (EC, c * @ wc ) is analogous to the product of concentration of soil water 
times volume of mobile soil water. Salt-free water is not a significant conductor of electricity; 
hence, the water in the soil is simply the "container" of the mobile electrolyte (the dissolved salt) 
and the "conduit" for the flow of electricity. Therefore, the effect that changes in water content 



FIGURE 15 


Relationship between slopes of EC e vs. EC a 


calibrations obtained for different soils of 


Montana, USA, and their clay contents (after 


Halvorson ef a/., 1977) 




■ i° y = io.astx.)" * 52 






12 


- \ r ■- O.B? 






_ _ _ 


i o Convention*! 






U10 

LlJ 


j, a Cell 
^° o EC -Pr*6* 






" 8 
hi 

% 6 

i/i 


\ 






2 








3 10 20 30 40 SO 60 7 





Clay content, C, % 





24 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



FIGURE 16 

Correlation between slopes of EC e vs. EC a 
calibrations obtained for different soils and 
their field-capacity water contents (after 
Rhoades, 1981) 



I0- 



..I 

S 




Sl O p*=-O.^TI(0 fL >* 12 2371 
t* =0.963 



have on EC a measurements and salinity 

appraisal depends on whether or not salt loss 

occurs with the change in water content. 

Immediately following an irrigation or rain 

event, salt removal from the soil occurs as the 

water content drains to "field capacity"; 

hence, measurements of EC a are relatively 

sensitive to changes in @ w during such times 

because the product (EC„ * @ w ) decreases in 

proportion to the change in water content. 

However, the soil is usually too wet during 

this period to permit one to access the field 

and to undertake measurements of EC a . These 

measurements generally only become feasible 

later and are typically made after the rapid 

drainage has ceased and the soil is at, or 

below, field capacity water content. During 

this latter period, further major losses of soil 

water in the rootzones of cropped soils occur 

mainly through transpiration and almost all of 

the salt contained in the water retained by the 

soil following irrigation is left behind in the 

remaining soil water; hence, the salt 

concentration (and EC„) of the remaining 

water is increased approximately 

proportionately to the reduction in @ w and, thus, the product (EC„ * @„) is approximately 

constant (hence, EC a ) is fairly independent of changes in water content following drainage. Thus, 

in saline soils, EC a is much more related to salinity (as expressed in terms of EC e or EC W ) than to 

water content and is, for any given irrigated and cropped soil, not much affected by the 

changes in water content that occur in the time period between the cessation of rapid drainage and 

the next irrigation event, i.e., during the period of time when measurements of EC a are normally 

made. The relatively small changes in EC a that do occur with changes in @ w under these 

conditions result from a variation in the partitioning of soil water between @ ws and @ wc 

(previously referred to in Rhoades et al. 1976, as a change in tortuosity) and from the 

precipitation of some salt as the solubilities of calcite and gypsum are exceeded. As @ w decreases 

below field capacity due to evapotranspiration, EC a will show a relatively small and 

approximately linear decrease according to the relationship: 



5 (0 15 SO 25 

Water CmUmi al field Capacirj-, © fc 



?,0 



AEC a = a A@ w K, 



[12] 



where a is a factor related to the relation between @ w c and @ w , , and K = EC„ @ w = a constant. 
For typical soils the error in EC a caused by a A@ w is not large with reasonable deviation in @ w 
from field capacity water content. Experimental evidence to support the above "argument" made 
on the basis of theory and logic have been obtained in both laboratory and field studies; some are 
reported in Rhoades et al., 1981, 1989a, 1989c, 1990a; Rhoades and Corwin, 1990; Rhoades, 



Soil salinity assessment 



25 



FIGURE 17 

(A) Cylinder and surrounding "moat" with impounded saline water used to leach the soil and 
adjust it to a desired level of salinity; (B) access-hole being made in soil with Oakfield-type 
soil sampling tube for subsequent insertion of EC a -probe; (C) EC a -probe being inserted into 
salinity-adjusted soil for determination of EC a ; and (D) sample of salinized soil being 
collected for subsequent determination of EC e (salinity) (after Rhoades et al., 1977) 







.- - , " : '-■<--■' rr- ■ - - - 


" 


■^ 


m 




I ^^p 


M 


^H 




Pb. 




\i . . .■ ■ 






26 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



1990b; Bottraud and Rhoades, 
1985b). Example data supporting 
this conclusion are given in Figure 
22, after Rhoades et at. (1981). 
Indirect evidence supporting this 
conclusion are the low correlations 
typically found between EC a and 
@„ in any given irrigated, salt- 
affected field but for which high 
EC, - EC e correlations are found. 
Johnston (1994) concluded in a test 
of this conclusion that the 
"compensation" described above 
was valid, but less than what the 
author found. However, he did not 
carry out his study in cropped fields 
where the "drying" mechanism is 
water removal by roots that are 
distributed throughout the entire 
relatively large volume of soil. He 
subjected small soil columns to the 
drying action of air which would be 
expected to cause water (and salt) 
to flow to the ends of the column 
away from where the electrodes 
were located. This mass flow is 
analogous to that which causes the 
changes in water and salt contents 
that occur in soils during drainage 
that, as discussed earlier, result in 
changes in EC, because the 
product of (EC WC * @ wc ) is 
reduced; whereas, changes in water 
content caused by transpiration do 
not reduce this product except when 
salt solubilities are exceeded. 

The appropriateness of using 
@„ as a reference for water content 
and of the inappropriate-ness of 
using matric potential (such as 
tensiometer readings) in 

establishing EC e = f (EC,) 
calibrations is supported by the 
results of the laboratory column- 
studies of Bottraud and Rhoades 
(1985b). These data are not 
reviewed here. Suffice it to say that 
EC, is directly related to @ w in the 



FIGURE 18 

Soil-filled, four-electrode cell (as obtained with a 
coring device) showing one group of four of the eight 
electrodes inserted into the undisturbed soil used to 
measure EC a ; after the soil is removed, it is analysed 
in the laboratory for EC e . (after Rhoades et al., 1977) 




FIGURE 19 


Relationship between bulk soil electrical conductivity 


and electrical conductivity of the saturated-paste 


extract for Dateland soil at field capacity water 


content (after Rhoades, 1980) 


H1I J I I J 1 J 1 




1 

• 
(J 
u 20 - 


Ofltelond SL A 

* / 

EC# = 8.5401^) -L42 / 

f 1 10.985 / 






a 

I! 

LU 

J 15- 

'*— 
a 

B 

o 
(ft 




- 




l » 1 

Q 

Ul 


/ 


- 




ft - 


X 






u \ i i i i i — r 

0.0 0.3 JO 1,5 2.0 2.5 3.0 3.5 




Elaetrtea! CftiHluetivily of Soil . EC^r dS/m 





Soil salinity assessment 



27 



manner previously described (and not 
to matric potential by any general 
relation) and if matric potential were 
used as a reference one would also 
need to know whether the soil was in 
a wetting or drying cycle. 

As shown above and in the 
previous section, EC a is highly 
influenced by salinity and, for any 
given soil, is not much influenced by 
normal variations in water content 
encountered during practical mea- 
surement times. However, as shown 
in equations [9] and [10], for a given 
salinity, EC a increases as @ fc , SP 
(which itself increases with clay and 
organic matter contents) and p b 
increase, because of their effects on 
the slope term, and as EC S , © s , and 
@ ws increase, because of their effects 
on the intercept term. EC S will 
increase as the clay content, cation 
exchange capacity and organic matter 
content of the soil increase. @ w , will 
increase with increases in clay 
content, organic matter content and 
bulk density, which itself generally 
decreases with increases in clay and 
organic matter. Thus, it is evident 
that soil texture and organic matter 
content, and correlated soil 
properties, will influence EC a and, in 
the absence of salinity, can be 
expected to be capable of being 
determined from sensor measure- 
ments of EC a , or EC,*, so long as the 
soils contain enough water to provide 
a continuous pathway for electrical 
current flow. The use of the 
mobilized sensor-surveys / "stochas- 
tic-calibration" approach described 
later is a very practical, efficient and 
accurate methodology for esta- 
blishing such EC a - soil property 
correlations and for developing much 
of the soil-property information 
required for prescription farming 
purposes (Rhoades etal., 1997d). 



FIGURE 20 

Relationships between bulk soil electrical 
conductivity and soil water electrical conductivity 
for the major soils of the Wellton-Mohawk Irrigation 
Project of Arizona USA (after Rhoades, 1980) 




01 2345676 
Electrical Conductivity at Soil ,£C f ,dS/m 



FIGURE 21 

Relationship between the soil electrical conductivity 
of Fallbrook soil and the electrical conductivity and 
sodium adsorption ratio (SAR) of the soil water (after 
Bottraud and Rhoades, 1985) 



Fallbrook Soil 



60 



■°. sua 
™ 
id 

- 4.0 

j; 
15 
>■ 3,0 

°> 
| 

o 
o 

| lit) 

5 









-l r 


— i — i — 


T T 




5AR 












V 


-0 










l* 


■ n 


= 10 






s 


*1 y 




□ 


= 20 






?' > 


-&/ 




o 


= 30 






s'X 






■ A 


■ 40 






,'' fm, <' 




- 


> 


= S0 




J* 


y^^J * 






• 


= C0 




/ , 


*%/ 










<4 


<*>/ 


/ 






- 


<f 


* s 


/ 






- 






s 
























£ j 












" 










































-" ■ 


■ 


1 


i i 


i ■ 


— 1 L 





Z.O AXt 6.0 SO IOJ0 ISO 14.0 ISO ISO zoo 
Elscirical Conductivity of Soil Water EC^ dS/m 



28 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



FIGURE 22 

Response of soil electrical conductivity to irrigations and evapotranspiration over several 

irrigation cycles (after Rhoades et al., 1981) 



o 



1 



E 
"-. 



o 



4 - 



H 



Soil Dtptli r m 

I.M 




_i_L 



j. 



276 £32 

Time. Julian Days 



309 



This prescription farming application is discussed more later. 

An equation analogous to that of [5] established for bulk soil electrical conductivity has 
been developed (Rhoades et al. 1989b) for saturated soil-pastes, as follows: 



E C P = 



(® S +® KS ) 2 EC WS EC S 
(& S )EC WS +(Q WS )EC S 



+ (Q w -Q ws )EC e 



[13] 



where the EC of the equilibrated extracted solution (EC e ) is analogous to EC WC , EC P is the 
electrical conductivity of the saturated-paste, @ w and @ s are the volume fractions of total water 
and solids in the paste, respectively, @ lvs is the volume fraction of water in the paste that is 
coupled with the solid phase to provide a series-coupled electrical pathway through the paste, EC S 
is the average specific electrical conductivity of the solid particles, and the difference (@ lv - ® ws ) 
is @„ c , which is the volume fraction of water in the paste that provides a continuous pathway for 
electrical current flow through the paste (a parallel pathway to @„ s ). Assuming the average 
particle density (p s ) of mineral soils to be 2.65 g/cm 3 and the density of saturated soil-paste 
extracts (p w ) to be 1.00, @ l( and @ s for saturated pastes can be directly determined from SP as 
follows: 



e„ =sp/[( Pw \oo/ Ps )+sp], 

and 



[14] 



S = 1 - 0w ■ 



[15] 



Soil salinity assessment 



29 



The saturation percentage of most 
mineral soils can be adequately 
estimated in the field, for purposes of 
salinity appraisal, from the weight of a 
known volume of paste (Rhoades et al. 
1989b). Figure 23 may be used for this 
purpose; for details of the relationships 
inherent in this figure see Wilcox 
(195 1). Evidence of the validity of this is 
shown in Figure 24. 

These relationships can be used to 
determine soil salinity using soil 
samples. The method requires the 
creation of saturated soil-pastes but 
avoids the need for the collection of the 
extract. Calibrations are needed for each 
different soil, but they are easily and 
accurately predicted by the means 
described in the next section. The 
method is faster and more field-practical 
than the conventional extraction 
procedures. 



Determining Soil Salinity from 
Saturated Soil-Paste Electrical 

Conductivity 

EC e can be determined from 
measurements of EC P and SP (using 
equations [13] to [15]), if values of p s , 
@ws and EC S are known. These 
parameters can be adequately and 
simply estimated, as demonstrated by 
PJioades et al. (1989b & c). For typical 
arid land soils of the Southwestern 
United States, p s may be assumed to be 
2.65 g/cm ; EC S may be estimated from 
SP as: EG = 0.019 (SP) - 0.434 (see 
Figure 25), and the difference (@ w - @ws) 
may be estimated from SP as: (@ w - @ ws ) 
= 0.0237 (SP)° ■"" (see Figure 26). The 
measurement of EC P and SP (from the 
volume-weight of the paste and Figure 
23) can be easily made using an EC-cup 
of known geometry and volume, 
conductance meter and battery operated 
balance as shown in Figure 27. This 
permits EG to be determined from 



FIGURE 23 

Theoretical relation between saturation percent- 
age (SP) and weight (in grams) of 50 cm 3 of 
saturated paste, assuming a particle density of 
2.65 g/cm 3 (after Wilcox, 1951 and Rhoades et al., 
1989b) 



IfiO 
10 



:■ 



50 



s 



=£ SO 
LU 

B 



E i0 



§ 



a: 

2 £0 



| ft I I 1 I I I I | I ■ I 1 | I I I ■ ) M I I | I I » l _| 



IV :□ iir. J VDUUNE 




qI ■ ' J ■ I ■ ■ ■ ■ [ - ■ ■ ■ I I ■ ■ I ■ ■ ■ ■ I ' 



SO 



TO 



BQ 



ll'J 



ISO 



WflMS TOSTE 



FIGURE 24 


Correspondence between measured and 


estimated (using Figure 23) saturation 


percentages, for a set of California soils (after 
Rhoades era/., 1989b) 


» " 








i i i i r i i 




« 


- 








mini IT'ttiT) > UWnl SI Jj 






C Tp 


. '••:)3 |>M ~flu 






u 

u 

ft, 








c id 

a 

3 

n SO 
Vt 


n-2t OT 






■o 

9 

r, C 

i 
■ 

t 


J J L 1 I 1 1 






1 £0 40 *0 6 


U 


Estimated Saturation Percentage. SP 





30 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



equations [13] to [15] (as 
described below, or from 
Figure 28) from the simply 
made measurements of the 
volume-weight and the EC of 
the saturated soil-paste. Evi- 
dence of the validity of this is 
shown in Figures 29 and 30. 
The method is suitable for 
both laboratory and field 
applications, especially the 
latter, because the apparatus 
is inexpensive, simple and 
rugged and because the 
determination of EC P can be 
made much more quickly than 
with the conventional 
procedure which involves the 
vacuum-extraction of the 
paste and, subsequently, the 
measurement of the EC of the 
extract. An additional savings 
in time occurs because less 
soil needs to be sampled (100 
grams is sufficient, which 
can be collected relatively 
quickly with a Lord-type, or 
similar, sampling tube) and 
less saturated-paste needs to 
be prepared (it takes about 
1/3 the time to prepare a 
saturated-paste for this 
method compared to the 
conventional method which 
requires about 400 grams 
and collection with a more 
time-consuming soil auger). 

As indicated above, 
ECe may be estimated from 
Figure 28 given EC P and SP, 
using the curve corresponding 
to the appropriate SP value, 
or else it may be calculated 
using the following equation: 



FIGURE 25 


Correlation between the electrical conductivity of the soil 


solid phase (EC S ) and the saturation percentage (SP) for 


some soils of the San Joaquin Valley of California USA (after 


Rhoades et al., 1989a) 


Q 


i i ■ ■ ■ i i 

EC, 10.01 9 SP- 0.434 




■S p 0.6 


L r'- 0.993 P 




>, -"v 


jT 




£ w 


tyr 




M ™ 0.6 


S 




M4 


/ 




H - 0.4 


j£ 




<J * 


jr 




HI 


j^ 




-£ O- 0.2 


/ 




y 


ty 




w 0, 






) 10 20 30 40 50 60 TO SO 


Sat urn Hon Percentage, sp ■ e e 



FIGURE 26 

Relationship between the volumetric content of water in the 
saturated soil-paste which is in the continuous electrical 
conduction path (@ w - @ws) and the saturation percentage 
(SP) for some San Joaquin Valley California USA soils (after 
Rhoades et al., 1989b) 



;■ Wf 



I M 



o* 



oa- 



02 



a i 



* o. 



T~ 



"f 



i" 1 1 1 1 r 

South King» River watershed Soils 



i8 u Q m J -00237 (SP) 
r* "O.BB 




an 



Saiuroi^fl fttrevibgff, SP 7 B e 



EC = 



-b+^b 2 -4ac 



2 a 



[16] 



Soil salinity assessment 



31 



FIGURE 27 

(A) Portable balance used in the field to determine the weight of the saturated soil-paste 

filling the "Bureau of Soils Cup", (B) "Bureau of Soils Cup" filled with saturated soil-paste 

connected to conductance meter, and (C) close up of "Bureau of Soils Cup" (after Rhoades 

1992) 




where a = [@ s (©» - © »,)], b = [(0, + @ws)* (EG) + (©» - 0» s ) (0» s EG) - (0 S ) EC P ], and c = - 
[(@ws) (EG) (EC P )]. The values of EG, ©s, @» and @ ws are estimated from SP using the 
relationships described above. 



Sensitivity analyses and tests have shown that the estimates used in this method are 
generally adequate for purposes of salinity appraisal of typical mineral arid-land soils (PJioades 
et al. 1989c). This method has been found to be quite accurate and robust (considerable 



32 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



experience and data have been obtained 
using it in the salinity assessment research 
of the US Salinity Laboratory). As 
discussed earlier, this method is a better 
choice to estimate EC e than those 
calculated from 1:5, and similar, soil:water 
extracts. For organic soils, or soils of very 
different mineralogy or magnetic 
properties, these estimates may be 
inappropriate. For such soils, appropriate 
values for @ s , EC S and @ ws will need to be 
determined using analogous techniques to 
those of Rhoades et al. (1989b). The 
accuracy requirements of these estimates 
may be evaluated using the approaches 
given in Rhoades et al. (1989c). 

It should be noted that (EC e @ e ) is 
not equivalent to (EC @w) because 
different amounts of soil are involved in 
the two measurements. The relationship 
between these two products is: 

EC„ e» /p b = ECe 0e/pp.[17] 

Data to support this is given in 
Rhoades (1981) and Rhoades et al. (1990). 
The ratio ®J p p is equivalent to SP/100 
(see Rhoades et al. 1989a, b). 

The procedure described in this 
section is especially suitable to determine 
the ECe values of the soil samples used to 
calibrate the stochastic-model (described 
later) which, in turn, is used to predict soil 
salinity from sensor measurements of Ed. 

This paste-method of determining 
ECe has been commercialized in the United 
States (see Table 12, Chapter 5) and is 
available in a kit form, including software 
for the calculations. This company also 
sells an analogous field-kit to determine 
soil sodicity (in terms of SAR e ) without the 
need for collecting extracts or performing 
analyses of calcium and magnesium 
concentrations, based on the methodology 
of Rhoades et al., (1997c). The sodicity 
method, while not as accurate as the 
salinity method, is sufficiently accurate for 



FIGURE 28 

Relationships between EC P , ECe and SP for 
representative arid-land soils (after Rhoades ef 
al, 1989b) 



■a 



U 

w 



111 
I 

1= 
2 

re 



m 



a 







i> 2 4 6 6 <S 12 '1 IS IB K> 



Eleclncal Conductivity of Setursjtitm- 



FIGURE 29 


Relationship between EC P and ECe for 


Grangeville soil. The symbols represent 


empirical data and the solid line is the "fit" of 


these data using Eq. [13] (after Rhoades et al., 


1989b) 

* 
4 

m 






1 1 1 i i /I 






rk 

u 

H 

•:' t 

£. 








i 


_j( GnMQnrilt 






1 * 


/ IkCi IS 

/ * **' 















" e 


/ r" -,M 






1 


/ * \'iVSt unao,:, miljl 






I 








C & K> IS 2d ZB SO 




Emmet! C«4icn«lf orSfhnlnn'EiiMKl.ECundi/iii 





Soil salinity assessment 



33 



many field diagnosis purposes and, most 
certainly, for screening samples to 
identify those that merit the time, labour 
and expense of laboratory analyses. 
These two kits permit soil salinity and 
sodicity to be determined directly in the 
field using saturated-pastes of soil 
samples. 



FIGURE 30 

Correspondence between measured and 
estimated (using Figure 29) soil salinities (EC e ) 
for representative soils of the San Joaquin Valley 
of California USA (after Rhoades et al., 1989b) 

10 







ft m 



I eo 



s 

Vt 

1 
1 



40 - 



ZQ 



determining soil salinity from 
Bulk Soil Electrical Conductivity 

Soil salinity can be determined from 
measurements of bulk soil electrical 
conductivity using essentially three 
different approaches. After reviewing the 
various instrumental means of measuring 
EC a , this section discusses these 
alternative methods of salinity appraisal. 

Sensors and equipment for measuring 
soil electrical conductivity 

Three types of sensors are commercially 
available for measuring bulk soil 
electrical conductivity. Two are field- 
proven, portable sensors: (i) four- 
electrode sensors and (ii) electro- 
magnetic induction sensors. A third sensor-type, based on time-domain-reflectometry (TDR) 
technology, has not yet been shown to be sufficiently accurate, simple, robust or fast enough for 
the general needs of field salinity assessment. Each of the first two sensor- types has its' own 
advantages and limitations. The first two sensors and related equipment will now be described 
and discussed. 

Four-electrode Sensors 



m*H EC, • d07 S +' LOOr'laU E£,'j / 






III40 (y/ 






f?",MG / 




_ 


& tt - I.OI6 QS 






of 




"' 


dfl 




- 


<sfe 




- 


f 1 1 1 L 1 


I J 


- 



SO 40 fO 90 

Eslimatsd Soil Sclintty, EC # , d&rtn 



Bulk soil electrical conductivity can be measured using four-electrodes inserted into the soil, a 
combination electric-current generator/resistance, or conductance meter, and connecting wire. A 
photograph of the basic "surface- array" equipment is provided in Figure 31. With such 
equipment the depth and volume of measurement may be varied by changing the spacing between 
the current (outside) electrodes, as illustrated in Figure 32. When the distance between the outside 
pair of electrodes (the current electrodes) is small, the flow of electricity is shallower than when 
the distance is greater. The effective depth of measurement is about one-third of the distance 
between current electrodes (see Figure 33). The spacing-depth relation is discussed in more detail 
later. The calculation of Ed from surface-array measurements requires knowledge of the spacing 
between the current and potential (inner pair) electrodes. An equation for calculating the "cell 
constants" for different arrays and spacings of electrodes is given later. 



34 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



FIGURE 31 

Four electrodes positioned in a Wenner-type surface-array 
and a combination electrical generator and resistance 
meter (after Rhoades and Oster, 1986) 




Oh 



t 



The current source- 
meter unit may be either a 
hand-cranked (see Figure 31) 
or a battery-powered type 
(see Figure 34). Some of the 
available generator/ meters 
were designed for 

geophysical purposes and 
these generally read in ohms 
of resistance. If such units 
are to be used for purposes 
of general soil salinity 
assessment, they should 
measure from 0.1 to 1000 
ohms. A unit specifically 
developed for soil salinity 
appraisal is available from 
Martek Instruments . It is 
battery powered, allows the 
geometry constant to be set 
for different configurations 
of electrodes and reads out 
directly in terms of soil 
electrical conductivity (an 
early model is shown in 
Figure 34; it is also more 
portable (smaller/lighter) 
than the typical units built 
for geophysical prospecting 
purposes). One can build a 
simple, but adequate, 
generator/meter unit from 
components using the 
circuitry-schematic and 

parts-list provided in Austin 
and Rhoades (1979) and in 

Annex 2. Units which can data-log time and EC a measurements have also been developed for use 
with both hand-held four-electrode sensors and with a mobilized, tractor-mounted version of a 
"fixed-array" unit developed for making automated "on-the-go" measurements of bulk soil 
electrical conductivity. The latter unit and mobilized system are described below. 

Electrodes used in surface arrays of the type shown in Figures 3 1 and 34 can be made of 
stainless steel, copper, brass, or almost any other corrosion-resistant, conductive metal. Array 
electrode size is not critical, except that the electrode must be small enough to support itself when 
inserted to a depth of 5 cm or less. Electrodes 1.0 to 1.25 cm in diameter by 45 cm long are 
convenient for most measurement purposes, although smaller electrodes are preferred for 
determining Ed within soil depths of less than 30 cm. The effect of depth of insertion of the 
electrodes is discussed later; an equation useful in this regard is given in Annex 3. Any flexible, 
well-insulated, multi-stranded, 12 to 18 gauge wire is suitable for connecting the array-electrodes 
to the meter. 



FIGURE 32 

Schematic showing increased depth and volume of EC a 
measurement with increased C1-C2 electrode spacing. 
Effective depth of measurement is approximately equal to 
one-third of (C1 - C2). C stands for current-electrode and P 
stands for potential-measuring electrode (after 
Rhoades, 1976) 



Smgll 
C, P, P 2 C 4 




Soil salinity assessment 



35 



For hand-carried mapping or 
traverse work, it is convenient to 
mount the array-electrodes in a 
board with a handle (see Figure 34) 
so that soil resistance, or 
conductivity, measurements can be 
made relatively quickly. Such 
mounted-units are practical for 
current-electrode spacings of up to 
about two metres; switching devices 
have been developed to make it 
easy to switch the meter quickly 
between the different sets of 
electrodes (Rhoades, 1976). These 
"fixed-array" units save the time 
involved in spacing the electrodes 
and keep the "geometry factor" 
constant from one measurement site 
to another. 

A mobilized, tractor-mounted 
version of a "fixed-array" four- 
electrode unit has been developed 
for making automated "on-the-go" 
measurements of bulk soil electrical 
conductivity. 

Generator/meter/logger units which 
can data-log time and EC a 
measurements have also been 
developed for use with this 
mobilized equipment, as well as 
with hand-held four-electrode 
sensors. The mobilized four- 
electrode system also data-logs the 
associated locations in the field, as 
determined using global posi- 
tioning system (GPS) equipment. 
This system, shown in Figure 35, is 
capable of making both faster and 
wider-spaced readings than can be 
accomplished manually, while 
simultaneously providing the x, y 
coordinates of each measurement 
site. It is especially well suited for 
collecting detailed information 
about the variability of average root 
zone soil electrical conductivity 
within fields and of the various soil 
properties that can be inferred from 

ECa. 



FIGURE 33 

Variation of: (A) current density with depth in a plane 
mid-way between the current electrodes; (B) current 
density at unit depth as a function of current electrode 
separation (after Rhoades and Ingvalson, 1971) 




0.2* 0.5O QJ5 

Depth n/Eieeiradfl Sepornhon 



J,00 




i 2 14$4 

Currani Elttlrtide Separation /b«pfh 



FIGURE 34 

A" fixed-array" four-electrode apparatus and 

commercial generator/meter (after Rhoades, 1978) 




36 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



FIGURE 35 

A mobilized (tractor-mounted) "fixed-array" four- 
electrode system, with mast for GPS antenna (after 
Rhoades, 1992) 




A close up view of one of the 
electrodes used in this system is 
shown in Figure 36. The use of the 
latter equipment to assess soil salinity 
and related management is described 
in more detail later and elsewhere 
(Rhoades, 1992a, 1992b, 1993, 
1994, 1996b; Carter et ai, 1993; 
Rhoades et ai, 1997a, 1997b). The 
electrodes and generator/meter can be 
attached to the tool bar of almost any 
tractor, or they can be provided 
"together with a dedicated tool bar 
and weather proof container for the 
generator/meter system (as shown in 
Figure 35). This mobilized/ 
automated four-electrode system is 
commercially available from 
Agricultural Industrial Manufactur- 
ing, Inc. . A more recently commer- 
cial mobilized four-electrode unit, 
though less well suited for use in 
furrow-irrigated cropped fields, is 
available from Veris Technologies . 

A salinity probe, in which the 
four electrodes are incorporated into 
a shaft, was developed by Rhoades 
and van Schilfgaarde (1976). With 
this probe, EG, can be measured in 
small soil-volumes and at various 
depths in the soil profile. Convenient 
sized current source-meter units have 
been designed for use with the four- 
electrode salinity (Austin and 
Rhoades, 1979). Commercial ver- 
sions of both the four-electrode probe 
and "meter" are made by Martek 
Instruments , by Fijkelkamp Agri- 
search Equipment and by Elico 
Limited . The probe and meter sold 
by "Fijkelkamp" are essentially the 

same as those developed by Rhoades and collaborators. The units sold by Martek Instruments are 
improved versions. The newest version of the Martek SCT meter, which reads directly in EC a 
corrected to 25 °C and which incorporates a data logger and a timer is shown in Figure 37, along 
with the moulded, insertion salinity-probes (both standard- and micro-sizes) they sell. Versions of 



FIGURE 36 

(A) Close-up of the four fixed-array electrodes used 
on the mobilized tractor-mounted system, (B) 
insulators used to isolate sensor part of shank from 
the rest of the tractor, and (C) close up of 
replaceable pad at bottom of electrode (after Carter 
etal., 1993) 




Mention of trademark or proprietary products in this manuscript does not constitute a guarantee or 
warranty of the product by the Food and Agriculture Organization of the United Nations and does 
not imply its approval to the exclusion of other products that may also be suitable. 



Soil salinity assessment 



37 



FIGURE 37 

Two commercial four-electrode probes (small and 
standard sizes) and electrical generator-meter/data- 
logger (after Rhoades, 1992). 




the four-electrode probes could be, 
but have not yet been, developed that 
are suitable for incorporation into a 
mobilized, automated system; 
however they can presently be used 
to advantage, even if manually, for 
some detailed assessment purposes 
(such as characterizing the salinity 
patterns within seedbeds and through 
root zones). Burial- type four- 
electrode units (see Figure 38) 
suitable for monitoring applications 
are also available from Martek 
Instruments. Simpler units can be 
built as originally developed by 
Rhoades (1979). Information in this 
regard is given in the Annex 4. Other 
special-purpose cells have been built 
for measuring EC a of undisturbed 
soil cores or in laboratory soil 
columns/cells. These units are shown 
in Annex 5. 

Electromagnetic-induction Sensors 

Soil electrical conductivity can be 
measured remotely using electro- 
magnetic induction (EM) method- 
ology. The basic principle of 
operation of the EM soil electrical 
conductivity meter is shown 
schematically in Figure 39. An EM 
transmitter coil located in one end of 
the instrument induces circular eddy- 
current loops in the soil. The 
magnitude of these loops is directly 
proportional to the electrical 
conductivity of the soil in the vicinity 

of that loop. Each current loop generates a secondary electromagnetic field that is proportional to 
the value of the current flowing within the loop. A fraction of the secondary induced 
electromagnetic field from each loop is intercepted by the receiver coil of the instrument and the 
sum of these signals is amplified and formed into an output voltage which is linearly related to 
depth-weighted soil electrical conductivity, Ed . The nature of the depth weighting is discussed 
later. 



FIGURE 38 

Commercial burial-type four-electrode conductivity 
probe used for monitoring changes in soil electrical 
conductivity (after Rhoades and Corwin, 1984) 




Figure 40 shows a commercially available EM soil salinity sensor (Geonics EM-38 ) 
oriented in both the horizontal (EMh ; Figure 40A) and vertical (EMv ; Figure 40B) coil-positions. 
This device was designed, at the request of the author, to meet the general-purpose needs of soil 
salinity appraisal (McNeill, 1992). The EM-38 device contains appropriate circuitry to minimize 
instrument response to the magnetic susceptibility of the soil and to maximize response to 
electrical conductivity. It has an inter-coil spacing of 1 metre, operates at a frequency of 13.2 



38 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



kHz, is powered by a 9 volt battery, 
and reads out directly in terms of 
EC a . The coil configuration, 
frequency and inter-coil spacing were 
chosen to permit measurement of 
ECa to effective depths of approxi- 
mately 1 and 2 metres when placed at 
ground level in horizontal and 
vertical configurations, respectively. 
Other "EM" units are available which 
are capable of deeper measurement. 
For more details about the principles 
of EM measurements and the various 
sensors that can be used in this 
regard, see McNeill (1980 and 1992). 

Mobilized, automated EM- 
measurements can be made within 
various depths of the rootzone using 
the EM-38 sensor, as well as with a 
combined four-electrode sensor if 
desired, and the mobilizing/ auto- 
mating equipment developed by 
Rhoades and collaborators shown in 
Figure 41 (Rhoades, 1992a, 1992b, 
1993, 1994, 1996b, Carter et al, 
1993). With this system, some 52 
operator-actions are automatically 
performed to collect a sequence of 
EM-38, four-electrode, and GPS 
readings at a given site. These 
actions are made in less than one 
minute (about 20 sites per hour, 
including travel time, can be sampled 
with this system). With these data, 
the salinity level and distribution 
within the soil profile to be 
determined in two dimensions. The 
EM-38 is contained within the 
cylinder protruding in front of the 
mobilizing unit. The sensor is in the 
"up" (travelling) position in this 
picture. The optional four-electrode 
component is shown in Figure 42 in 
the "down", inserted position. As 
with the mobile, four-electrode 
system previously described, this 
system also incorporates synchro- 
nized, GPS site-positioning equip- 
ment and data logging capabilities. 



FIGURE 39 

Schematic showing the principle of operation of an 

electromagnetic induction soil conductivity sensor 

(after McNeill, 1980) 




INDUCED CURRENT fLOW IN GHOWiD 



FIGURE 40 

Geonics EM-38 electromagnetic soil conductivity 
sensor in (A) horizontal orientation and (B) vertical 
orientation (after Rhoades, 1992) 




B 



FIGURE 41 

Mobilized salinity assessment system with 
combined EM-38 and four-electrode soil 
conductivity sensors and mast for mounting a GPS 
antenna. Both sensors are in the 'up' travel position 
(after Rhoades, 1992; 1993; Carter et al., 1993) 




Soil salinity assessment 



39 



This mobilized/automated system is 
very well suited for the detailed 
mapping of EC a * and correlated soil 
properties, as well as for the mapping 
of these properties within different 
depth-intervals of the root zone and 
slightly deeper. The system is 
commercially available from AIM 
Inc. . It also incorporates a laser mast 
and a load-cell which permits the four- 
electrode array to also function as a 
penetrometer (not shown). Thus, 
additional, complementary informa- 
tion can be simultaneously obtained 
about the micro-relief features of the 
field and about the compaction and 
crusting properties of the soil surface. 
The latter co-located information helps 
determine the cause(s) of salinization, 
and the appropriateness of irrigation/ 
drainage management, considering the 
observed salinity levels and patterns. 
Several prototypes of this system have 
been designed and tested; they are 
described in more detail in Rhoades 
(1992a, 1992b), Rhoades, (1993), 
Carter et al. (1993), Rhoades (1994), 
Rhoades (1997b), and Rhoades et al. 
(1997a, 1997b, 1997d). Other forms of 
mobilization of EM-sensors have been 
undertaken, though they are not as 
integrated nor as well adapted to row- 
cropped fields as the above described 
system (Cameron et al., 1994; Jaynes, 
1996; Kitchen et al, 1996). 



FIGURE 42 

Close-up of the fixed-array four-electrode unit in 
(A) the travel-position and (B) inserted into the soil 
by the hydraulic "scissors" apparatus of the 
mobilized combination sensor assessment system 
(after Carter et al., 1993) 




Procedures for Measuring Bulk Soil Electrical Conductivity 

Large-volume Measurements 

For the purpose of determining soil salinity within root zones, or some fraction thereof, it is 
desirable to make the measurement of EC a to depths of up to 1 to 1.5 metres. This may be 
accomplished with both four-electrode and EM-sensors. It is accomplished with the four-electrode 
equipment by configuring the surface-array of electrodes in a straight line with the spacing 
between the two outer (current) electrodes selected depending upon the desired depth(s). As 
implied in Figure 32, the depth and volume of measurements are readily altered by varying the 
spacing between the current-electrodes. The relative spacing between the inner (potential- 
electrode pairs can also be varied, but this does not affect the depth of measurement. The 
electrodes are often spaced in the so-called Wenner-array with equal spacings between all of them 
(Wenner, 1916; Rhoades and Ingvalson, 1971). When using the Martek SCT meter, each of the 
inner-pair of electrodes is preferably placed inward from its closest outer-pair counterpart a 



40 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



distance equal to 10 % of the spacing between the outer-pair. In both of the above-mentioned 
electrode-arrangements, as well as for others, the inner-pair of electrodes is generally used to 
measure the electrical potential (or resistance) while current is passed between the outer-pair. The 
effective depth of current penetration for either configuration (in the absence of appreciable soil 
layering) is approximately equal to about one-third the outer-electrode spacing, y; thus "average" 
soil salinity is measured to a depth equal to approximately y/3 (Rhoades and Ingvalson, 1971; 
Rhoades, 1976; Halvorson and Rhoades, 1976). Thus, by varying the spacing between current 
electrodes, one can measure salinity to different depths, also of different volumes, in soil using the 
four-electrode system. 

An advantage of this "surface-array" method is the relatively large volume of soil that is 
measured compared to that of the insertion four-electrode probes (discussed later) or of 
customary soil samples. The volume of measurement is about (y/3) , where y is as defined above. 
Hence, effects of small-scale variations in field-soil salinity can be minimized by these relatively 
large- volume measurements. 

For measurements taken in the Wenner-array (electrodes equally spaced) using geophysical 
type meters, which measure resistance, bulk soil electrical conductivity is calculated, in dS/m, as: 



EC, = 159.2/, /a R, 



[18] 



where a is the distance between the electrodes in cm, R t is the measured resistance in ohms at the 
field temperature t, f, is the previously described temperature compensating factor used to adjust 
the reading of EC a to a reference temperature of 25° C, and 159.2 is the numerical equivalent of 
1000/2 %. For measurements made with other spacings of electrodes, EC a is calculated (after 
Dobrin, 1960) as: 



EC a =(\QQOI2%R) 



fj 



1 1 1 



R, R^ 



[19] 



where R is resistance, / ( is the 
temperature correction factor, and 
ri , r2, Ri and R? are the distances in 
cm between various pairs of 
electrodes (Figure 43). For 
measurements made with the 
Martek SCT meter, the meter is 
calibrated for any set of electrode 
spacings by setting it to read the 
value of ECa calculated from 
Equation [19] while the corres- 
ponding resistance is connected 
between the outer-electrodes. Use 
of a variable resistor box is 
convenient in this regard. 



FIGURE 43 

Schematic of distances between the current and 
potential electrodes in four-electrode array for use 
with Equation [19] (after Dobrin, 1960) 



J *f 



Rr 



-* R; 



Soil salinity assessment 41 



In both Equation [18] and [19], the electrodes are assumed to make only point contacts 
with the soil. In practice they must be inserted into the soil far enough to support their weight and 
to make contact with soil having sufficient moisture to permit a meaningful measure of EC a and 
interpretation of salinity. An equation is provided in Annex 3 to correct for this depth of insertion, 
assuming the soil is uniform throughout this depth and the rest of the soil volume involved in the 
measurement. Since this assumption is seldom true, the "correction" equation given in Annex 3 is 
seldom used in practice. Instead, an attempt is made to minimize the depth of insertion, especially 
for shallow soil depth measurements. Based on empirical findings of Rhoades and Ingvalson 
(1971), the depth of insertion should be no more than 25 mm for measurements within the 0-0.3 
m soil depth and no more than 50 mm for measurements within the 0-0.6 m soil depth. For deeper 
soil depth measurements, the electrodes may be inserted up to depths of 75 mm with no 
discernible effect. The diameter and length of the electrodes should be reduced when 
measurements are to be made with them inserted to shallow depths; otherwise, the larger 
electrodes can not be supported by the soil and make good contact. They tend to fall over and to 
"break contact" with the soil. 

In most field situations, the immediate topsoil is too dry and loose to attempt to make 
measurements when the electrodes are inserted to shallow depths. Hence, they are inserted to a 
depth below the boundary separating the dry/loose surface mulch and the underlying moist/firm 
soil, at whatever depth this boundary occurs. This depth is then taken as the "zero-depth" for 
inferring soil salinity from the Ed measurements, or for collecting soil samples to be used to 
establish EC e - EC a calibrations for the soil, since no useful relation exists between salinity and 
EG in dry soils as explained earlier. The fact that four-electrode measurements do not apply to 
the dry/loose surface mulch through which the electrodes were inserted in order to reach 
moist/firm soil may be seen as either a serious limitation of the surface-array, four-electrode 
method, compared to the EM method/sensor (Johnston, 1994), or as an advantage for reasons 
given later. Unfortunately, the incorporation of such surface soil, especially when it is highly 
salinized through evaporation-driven processes, in soil samples collected to calibrate or test four- 
electrode systems (also EM systems) has resulted in some erroneous calibrations, 
misinterpretations, and conclusions reported by a number of users and even investigators. It needs 
to be stressed that the variability that exists over very short distances in surface irrigated soils is 
often very great. Hence, great care must be made to avoid the collection of samples from regions 
of the soil that are not within the volume of the four-electrode sensor measurement when 
establishing calibrations or applying/testing them. A good example of such short-scale variability 
is shown in Figure 73, which is presented and discussed later. It needs to be recognized, as 
mentioned earlier, that what is many times used as "truth" regarding soil salinity in testing the 
models, data, various sensors, approaches and methodology involved in salinity assessment is, in 
fact, often not an adequate/appropriate index of salinity; certainly not that involving dry soil 
samples and high water:soil ratio extracts. Special care must be taken to account for the spatial 
variability that exists in typical saline soils. The mobilized measurement systems and stochastic- 
model approach of salinity assessment, which is discussed later, were explicitly developed to 
provide practical tools to measure and characterize spatially variable soil salinity in irrigated 
fields. Examples will be given later to illustrate the utility of these systems and approach to 
describe and account for this spatial-variability dilemma. 

Relatively large volumes of soil can also be measured with the EM-38 sensor, though less 
than with the tractor-mounted, fixed-array, four-electrode system. The volume and depth of EM- 
sensor measurements are influenced by the spacing between coils, the current frequency, and the 
orientation of the axes of the magnets/coils with respect to the soil surface plane (McNeill, 1980). 
The effective depths of measurement of the Geonics EM-38" device are about 1 and 2 metres 



42 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



FIGURE 44 

Cumulative relative contribution of all soil 
electrical conductivity, R(Z), below various 
depths sensed by the EM-38 unit when placed 
on the soil surface in horizontal (parallel) and 
vertical (perpendicular) magnetic-coil positions 
(after McNeill, 1980) 



when it is placed on, or in close proximity 
to, the ground and the coils are positioned 
horizontally and vertically, respectively. 
The effective width of the measurement 
extends out about Vi metre to the sides and 
ends of the unit (McNeill, 1990). Thus the 
elliptical volume of the measurement has 
an length of about 2 m, a width of about 1 
m and a depth that corresponds to the "Z' 
values given in Figure 44 and described 
by the equations provided in footnote 
number 2. The depth (and associated 
volume) does not contribute equally to the 
measurement, as explained next. 

The EM-38 device does not provide 
a linear measure of EC a with depth, rather 
a depth-weighted value Ed is obtained as 
stated earlier. The theoretical depth- 
distributions of this weighting for an 
homogeneous soil are shown in Figure 44 
for both the vertical (EMv ) and horizontal 
(EMh) configurations. The ratio of these 
distributions is shown in Figure 45. These 
distributions show that the EMh and EMv 
measurements are not independent; they 
are interrelated measurements, though not 
so much in the shallow depths (such as 0- 
0.30 m). The to 0.3, 0.3 to 0.6, 0.6 to 

0.9, and 0.9 to 1.2 m soil depth intervals contribute about 43, 21, 10, and 6 percent, respectively, 
to the EC a reading of the EM unit when it is positioned on homogeneous ground in the horizontal 
position (Rhoades and Corwin, 1981). Thus, the depth-weighted bulk soil electrical conductivity 
read by the EM device for this situation and in this configuration is approximately: 



If 


— r — i — i — r — i — i — i — l — P — ■ — •—> — P — I — 1 


0^ 


\ 


OH 


"\ \ 


07 


*\ \ 


06 


- \ \ 


OS 


- \ \ 


04 


\ X 


03 


\ ^N. 




^v ^-~ Vtolittfl 


CZ 


^ S ^-_ ^ _~ 


yi 


HofiiMk*" ~~ 


"CO Q2 04 0£ O.fl 10 IZ 14 l.fi 1.8 'iH 22 24 £6 28 ' 



DEPTH, 



ECa =0.43EC„ 



+ 0.21EC, 



+ 0.1EC a 



, + 0.06EC a 



+ 0.2EC a 



[20] 



where the subscript designates the depth interval in metres. Corresponding percentages in the 
vertical position are 17, 21, 14 and 10, respectively. Recent studies show that these proportions 
do not hold for non-homogeneous profiles (Rhoades et al. 1990b). The EMh and EMv 
measurements made with the EM-38 measurements also depart differently from the actual EC a 
values, even in homogeneous soils, when these levels exceed about 2 dS/m, as shown in Figure 
46. For these reasons, the "profiling" methods based on theoretical, uniform-profile weighting 
functions (such as that of Slavich, 1990) are not generally reliable in their applications. Methods 
that account for the effect of non-uniform weighting functions need to be used. Some of these 
methods are discussed/referenced later. 



The relative contributions (R) to the secondary EM field (or EC a ) from all material below a depth 



can be theoretically calculated from Rv = 1/(4" + 1) , and Rh 
and horizontal (H) dipoles, respectively (McNeill, 1980). 



(4 " + 1) - 2, for the vertical (V) 



Soil salinity assessment 



43 



FIGURE 45 

Ratio of vertical and horizontal weighted responses 
of the EM-38 unit as a function of compsite depth 
increments (i.e. 0-0.15, 0-0.30, 0-0.45, 0-0.60 m, etc.) 

(after Corwin and Rhoades, 1990) 



■a 
o 



p 



-I 



E 



Q 

1 




'0 02 Q4 QE OB ID \2 14 16 
COMPOSITE DEPTH INCREMENT^ (malres) 



The EM-38 does not lend itself 
well to the direct determination of 
average EC a (or EC e ) since its 
response is weighted by depth 
(generally decreasing with depth). 
However, as pointed out by Rhoades 
and Corwin (1981), the depth- 
weighted reading it provides 
(Equation 20]) is close to that which 
many researchers regard as the way 
crops extract soil water within their 
root zone and the way they 
proportionately respond to the 
variation of salinity with depth in the 
root zone. For this reason, Rhoades 
and Corwin (1981) suggested that the 
ECa reading obtained with the EM- 
38 might provide a reasonable 
measure of crop-effective salinity. 
Based on this concept/suggestion, 
Wollenhaupt et al. (1986) developed 
depth-weighted calibrations for the 
EM-38 and soils they worked with in 
Canada using salinity values by 
depth in the soils weighted in 
accordance with slightly modified 
versions of the depth-response 
relations given above. Subsequently, 
McKenzie et al. (1989) developed 
analogous but slightly different 
weighted-calibration relations for the 
EM-38 for use with their soils, as did 
Johnston (1994) in South Africa. 
There is some evidence that crop 
response to salinity can be reasonably 
related to the depth-weighted reading 
provided by the EM-38 (Slavich and 
Read, 1983; McKenzie et al, 1990; 
Rhoades et ah, 1997d). In spite of 
this, the latter three sets of 
calibrations referenced are subject to 
the same criticism given above about 
the inapplicability of the theoretical 
depth-response of the EM sensors 
(which are based on uniform-depth 

salinity conditions) to depth-varying conditions of soil salinity. However, they can be used 
advantageously for applications where relative differences in salinity need to be mapped in 
landscapes; but one should not expect to accurately determine the salinity distribution through the 
soil profile using such calibrations. 



FIGURE 46 

EM-38 readings for homogeneous profiles as a 

function of the profile EC a value (personal 

communication from J. D. McNeill; Rhoades, et al, 

1990) 




EMh ■ <L.»t]3(GCal - D.WKK. *] / 




| B0 


EM»-0.BW!(EtJ-Q.O1SS<E€,i) / X 




3 8.0 

> 
■ 

m 

1 « 

1 






1 2 " D 

V 

0.0 





■hor-zxinljQlrQnriffyrBtrift 












2 *0 6.0 3.0 10 


.0 



44 Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



It is often desirable to be able to determine soil Ed within various depth intervals so that 
soil salinity levels within the various parts of the root zone can be calculated, as needed for 
making certain assessments and management decisions. Since the proportional contribution of 
each soil depth interval to the EM-38 reading (Ed ) can be varied by changing the coil 
orientation or, as shown by Rhoades and Corwin (1981), by raising the unit to various heights 
above the ground, it is possible to estimate Ed within various depth-increments of the soil from a 
succession of EM measurements made at various orientations, or at various heights above- 
ground, or both. The Ed values within different discrete soil depth intervals of a group of 
California soils have been found to be correlated with a succession of EMh readings made above 
ground as: 

ECa, 0-0.3 = otoEMo + cciEMi + 0C2EM2 + 00EM3 + (X4EM4, [21a] 

EC, 0.3-0.6 = P0EM0 + (3iEMi + P2EM2 + p 3 EM 3 + p 4 EM 4 , [21b] 

where EM represents the reading obtained with the EM-38 unit held in the horizontal position and 
0, 1, 2, 3 and 4 represent height above ground in increments of 30-cm The author has found 
these relations to apply in each of the several widely distant places in the world where he has 
personally tested them. But since they have not been widely tested and may be expected to vary 
for different Ed - depth patterns, for the reasons described above, they should be used with 
caution until they have been evaluated for the specific conditions of interest. 

Subsequently, another series of empirical equations and coefficients were developed to 
estimate Ed within discrete soil depth intervals using just two measurements made with the 
magnetic coils of the EM-38 instrument positioned at ground level, first horizontally and then 
vertically (Corwin and Rhoades, 1982, 1984). These equations were initially developed using the 
same relatively small data set as that involved in the development of Equation [21]. Salinity 
increased with depth in all of these soil profiles. Subsequent tests showed the new empirical 
relations were inaccurate when applied to soils with salinities which decreased with depth; hence, 
a new set of coefficients were developed for such soils (Corwin and Rhoades, 1984). The two 
categories of soils were distinguished by the EMv / EMh ratio. Profiles in which salinity 
increased with depth were called "regular" profiles and were associated with EMv / EMh ratios 
of > 1 . Profiles in which Ed (salinity) decreased with depth were called "inverted" profiles and 
were associated with EMv / EMh ratios of < 1 . The derivation of these relations and the resulting 
general form of the equation are given in Annex 6. These relations were subsequently 
modified/improved using a substantially larger data base (Rhoades et al. 1989d) and expressed in 
the following form: 

(ECa, xi-xif 25 = fa (EMh) ' 25 + k v (EMv) ' 25 + k 3 , [22] 

where xl - x2 represents a given depth increment in cm, EMv and EMh are the readings obtained 
with the EM-38 device positioned at the soil surface in the vertical and horizontal positions, 
respectively, kg , kv and k3 are empirically determined coefficients for each depth increment, and 
the exponent 0.25 is an empirical factor used to provide a more normally-distributed set of 
values. This approach based on but two EM-38 readings is more practical to use than the 
approach inherent in Equation [21] which requires five measurements. Equation [22] is also more 
easily solved than is Equation [21] and the results were found to be almost as accurate for the 
two depth intervals 0-30 and 30 - 60 cm, when tested using the same original data set. Johnston 
(1994) evaluated these relations and those of Slavich (1990) for their applicability in South 
African soils; he found Equation [22] to be more accurate for the variety of soils and situations 



Soil salinity assessment 



45 



he tested. In fact, Johnston 
concluded that the agreement 
between his measurements and those 
predicted using the published 
relationships based on Equation [22] 
(given in Rhoades et al., 1989d) was 
"impressive", with "very little bias 
or error" and with very good 
correspondence (r = 0.89; slope = 
0.944, and intercept = 0.110). In 
contrast, he found the relationships 
of Slavich (1990), which are derived 
from the theoretical homogeneous 
depth-response functions (and as 
previously discussed are concluded 
to be inapplicable for non- 
homogeneous conditions), to have 
much poorer correspondence (slope 
of 0.71) and to produce a strong 
systematic prediction error. 



FIGURE 47 


Theoretical relation between In EMh and the difference 


(In EMh - In EMv) for uniform EC a profiles (after Rhoades, 
1992) 


3.0 


P 




1.5 


J / 




w 0.0 






-1.5 


1 




-0.2 0.0 0.2 0.4 


ln(EMHHn(EMv) 



Surprisingly, Johnston (1994) did not evaluate the following improved, more rigorous and 
more general relationship that was developed more recently and based on a larger and more 
varying set of EC a profiles (Rhoades, 1992a) and which the author concludes to be more 
generally applicable and accurate than those represented in Equation [22] : 



In ECa = po + [31 In EM H + p 3 (ln EM H - In EMv ), 



[23] 



where po, Pi and p? are empirical coefficients. In the earlier approach inherent in the 
development of Equation [22], two profile types were distinguished based on EMv / EMh ratios - 
regular (EMv > EMh ) and inverted (EMh > EMv ). Equation [22] and its manner of use have at 
least three deficiencies. The non-linearity that exists in the EMh - EC a and EMv - EC a 
relationships that occur at high values of EC a (see Figure 46, after Corwin and Rhoades, 1990) is 
not taken into account; near-uniform profiles are incorporated into either regular or inverted 
types, and the colinearity that exists between EMh and EMv (see Lesch et al. 1992) is not taken 
into account. Equation [23] minimizes these deficiencies by separating soil profile types into three 
classes (regular, uniform and inverted), by utilizing curvilinear EMh - EC a and EMv - EC a 
relationships to identify the three profile types, and by using the difference (In EMh - /n EMv ) in 
place of EMv as the second variable in the relationship, in order to minimize the colinearity 
problem. 

The theoretical relation between /n EMh and (/n EMh - /n EMv ) uniform EG, profiles is 
shown in Figure 47. The fitted curve ((In EM H - In EMv) = 0.04334 + 0.03058 In EM H + 
0.00836 EMh*)) describes a theoretically uniform Ed profile. Profile types may be classified 
based on deviation from this relation better than by the EMv / EMh ratio. For the practical 
purposes of solving Equation [23], the profile types have been classified as follows, after 
Rhoades (1992a): sites having values of (In EMh - In EMv) within ± 5% of the theoretical value 
(i.e., 0.04334 + 0.03058 In EM H + 0.00836 EM H 2 ) are designated "uniform"; those with 
measured values > 5% of the theoretical are designated "inverted", and those with measured 
values < 5% of the theoretical are designated "regular". Empirically determined values of the 



46 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



coefficients for Equation [23] based on these classification criteria and empirical data obtained 
from a large number and wide variety of California soils are given in Table 1. 

TABLE 1 

Relationships for predicting soil electrical conductivity within soil-depth intervals from EM-38 

readings 3 



Depth 
(cm) 


Predictive Equation 


n 


2 

r 


For Regular Profiles (measured values" < 5% of theoretical value) 


0-30 

30-60 

60-90 


In ECa = 0.414 + 0.985 In EMh + 2.336 (In EMh - In EMv) 
In ECa = 0.836 + 1.262 In EMh + 1.307 (In EMh - In EMv) 
In ECa = 0.674 + 1.089 In EMh - 0.446 (In EMh - In EMv) 


650 
626 
200 


0.76 
0.75 
0.69 


For Uniform Profiles (measured values" within 5% of theoretical value) 


0-30 

30-60 

60-90 


In ECa = 0.478 + 1 .209 In EMh + 0.41 1 (In EMh - In EMv ) 
In ECa = 0.699 + 1.234 In EMh - 0.623 (In EMh - In EMv) 
In ECa = 0.477 + 1.053 In EMh - 0.691 (In EMh - In EMv) 


73 
70 
24 


0.81 
0.81 
0.81 


For Inverted Profiles (measured values" > 5% of theoretical values) 


0-30 

30-60 

60-90 


In ECa = 0.626 + 1.239 In EMh + 0.325 (In EMh - In EMv) 
In ECa = 0.881 + 1.216 In EMh - 1.318 (In EMh - In EMv) 
In ECa = 0.563 + 1.206 In EMh - 1.641 (In EMh - In EMv) 


56 
55 
21 


0.91 
0.81 
0.91 



Predictions are based on measurements made with the EM-38 sensor placed on the ground in the 
horizontal (EMh) and vertical (EMv) configurations. Comparing measured values of (h EMh - h 
EMv) with the theoretical value = (0.04334 + 0.03058 In EMh + 0.00836 EMh 2 ). 



As mentioned earlier, the immediate topsoil is often dry and loose when measurements of 
ECa need to be made. This has been viewed by some as a "problem" and disadvantage of the 
four-electrode method compared to the EM method, because of the poor electrode contact that 
occurs with the former method under such conditions and the lack of need for contact with the 
latter method (Johnston, 1994). However, as explained earlier, no useful information about 
salinity can be made from EC a measurements made upon dry soil (dry soil behaves essentially 
as an insulator). Hence, when these measurements are made upon a soil having dry/loose surface 
mulch, the electrodes should be pushed through the "insulating" depth and into the moist soil to a 
depth of 25-75 mm (the minimum required for the electrode spacing, as explained earlier). In this 
manner, the dry/loose mulch does not affect or enter into the measurement of Ed; hence, no 
attempt should be made to infer salinity in the depth of "bypassed" soil, nor to include it in any 
calibration relation. When the procedure is followed as just explained, which has always been the 
case with the author, there is no "problem" or disadvantage in the use of the four-electrode sensor 
compared to the EM-38 sensor. To the contrary, the "problem" and limitation more often occurs 
with use of the EM sensor. Because no contact is required with this sensor, many users simply 
place the unit on top of the dry/loose soil and read and interpret/calibrate it as if the "insulating" 
layer contributes to the reading. Thus, the EM-38 reading made in this manner includes an error, 
especially the EMh reading, that is proportional to the depth of the dry layer times its weighting 
contribution. If the dry layer of soil is included in the calibration of the EM unit it will be in error; 
if the user attempts to infer the salinity in this layer from the EM-reading, it will be in error. It is 
inappropriate to include the dry layer in a depth-increment sample that includes moist soil. The 
errors created can be substantial, especially where salts are concentrated in the solid form by 
evaporation in the near-surface soil. Whenever feasible, the dry/loose soil should be scraped away 
from the measurement site before positioning the EM-38 sensor and measuring EG, . This has 
been the routine practice of the author and his collaborators and is included in the calibrations 
that they have reported. It takes more time and effort to use the EM sensor in this more 
appropriate manner, compared to the four-electrode unit, because the dry/loose soil must be 
removed for the former sensor but not the latter unit. With the four-electrode unit you simply 



Soil salinity assessment 47 



push the electrodes through the dry/loose layer; one can easily feel when moist soil is encountered 
and the depth of effective insertion is simply referenced from this point. 

Another practice routinely used by the author and collaborators involving the EM-38 for 
about the last 6-7 years, is to position the axis-centers of the magnetic coils 100 mm above the 
soil surface at the time EC a measurements are made. This practice is included in the most recent 
calibrations reported by Rhoades (1992a). This practice was instituted for several reasons. One is 
the observation made in the study of the EM-38 response to depth-varying Ed distributions 
(Rhoades et al., 1990b) that the reading of EMh was higher (near a maximum) when the sensor 
was held at a height of 100 mm above ground than when placed on the ground. When questioned 
about the reason for this phenomenon, the manufacturer (McNeill, 1990) could offer no physical 
explanation for this phenomenon but confirmed that he had also observed it. Another reason was 
to keep the instrument clean by avoiding contact with muddy or dusty soil. A device was built to 
permit the sensor to be positioned so that the coil- axis would be located 100 mm above the soil 
during both EMh and EMv measurements. It is shown in Annex 7. This device is also used 
advantageously to scrape away the dry/loose surface soil before it is positioned on the ground; it 
is also used to "level" the EM-38 sensor. The EM-38 readings will vary some as the sensor is 
tilted with either the transmitter or receiver end held higher than the other. The third reason was 
that some height was needed to clear the clods and rough surface that exist in many fields when 
the sensor was incorporated into the mobilized system of measurement described earlier (Figure 
41). The advantage of measurements taken at a height of 100 mm is that essentially maximum 
readings (depths of signal "penetration") can be achieved while avoiding contact of the sensor 
with the soil. It is not practical to remove the dry/loose surface layer of soil when EM-38 
measurements are made with the automated/mobilized system described earlier. The associated 
error is minimized using a stochastic, field calibration method that creates calibrations for the 
specific field conditions and methods of measurement. This stochastic-method is explained later. 

Small-volume Measurements 

Sometimes information is desired about the levels of salinity within small, localized volumes of 
the soil, such as that within different sections of the seedbed or under the furrows, and about the 
distribution of salinity within the rootzone. For such uses/needs, the insertion four-electrode EC- 
probe developed by Rhoades and van Schilfgaarde (1976), and commercialized by Martek 
Instruments , Eijkelkamp Agrisearch Equipment and Elico Limited , is recommended. The 
insertion EC-probe available from Eijkelkamp Agrisearch Equipment is a mechanically 
constructed unit consisting of four annular-ring electrodes separated by insulators directly 
patterned after that of Rhoades and van Schilfgaarde (1976). The construction details of the latter 
probe are given in the Annex 8 for those who might wish to construct their own. In the standard- 
sized "Martek" probe (see Figure 37), the four annular electrode-rings are molded into a plastic 
matrix that is slightly tapered so that it can be inserted into a hole made to the desired depth with 
a Lord - or Oakfield -type coring tube (or one of similar diameter, -25 mm). The smaller-sized 
probe (so-called "bedding" probe, see Figure 37), can be simply pushed into the soft upper-soil to 
the desired depth. In either sized unit, the probe is attached to a shaft (handle) through which the 
electrical leads are passed and connected to a meter. Burial type units are also commercially 
available in which the leads from the probe are brought to the soil surface (see Figure 38). The 
original burial unit developed by Rhoades (1979b) is simple and cheap to construct, as shown in 
Annex 4. A multiple-depth version of the four-electrode probe has been built by Nadler et al. 
(1982). The volume of sample under measurement with any of these probe-sensors can be varied 
by changing the spacing between the current electrodes and the over-all diameter of the probe. 
The standard-sized, Martek SCT Probe, has a spacing of 6.5 cm between outside electrodes and 



48 Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



measures a soil volume of about 2350 cm . The Martek "bedding" probe measures a soil volume 
of about 25 cm . Other four-electrode cells and units have been designed for other purposes 
including: making measurements on undisturbed soil-cores (Rhoades et al., 1977), making 
measurements at variable water contents (Rhoades et al., 1976; Bottraud and Rhoades, 1985a, 
1985b; Johnston, 1994) and making measurements in soil columns (Shainberg et ah, 1980; 
Bottraud and Rhoades, 1985b). Examples of some of these units are shown in Annex 5. Very 
small, fixed-array units have also been constructed to measure Ed along the wall of exposed soil 
profiles and in very shallow depths of soil surfaces (an example is shown in Annex 5). The 
possibilities are numerous and four-electrode units are easily made for such specialty 
purposes/studies. 

When using meters which display resistance, Ed in dS/m is calculated, for any of these 
probes, as: 

EC, =kfi/R t , [24] 

where k is an empirically determined geometry constant (cell constant) established for the probe 
in units of 1000 cm , R is the resistance in ohms at the field temperature, and/ is the factor used 
to adjust the reading to a reference temperature of 25° C (see Equation [2]). With the Martek 
unit, values of Ed are given either at field temperature or at 25° C. Since the time response of 
this thermistor is slow, because it is embodied in the probe, it is usually more convenient to use 
an electronic temperature sensor to measure soil temperature. The author has used a soil 
temperature probe sold by the Wahl Company for more than ten years with very good results. In 
normal field applications, Ed readings are generally made in the uncorrected temperature mode 
and the temperature distribution throughout several soil profiles in the field is determined either 
with the temperature read-out of the Martek unit or, preferably, with a faster electronic 
temperature sensor. Subsequently the Ed readings made at field temperature are converted to 
25° C values using Equations [2] and [24]. 

Procedures for Interpreting Soil Salinity 

Soil salinity, in terms of either Ed or Ed , can be determined in the field from measurements of 
bulk soil electrical conductivity by essentially one of three ways. Each has its own advantages 
and disadvantages. These alternative ways will now be described. 

"Specific Field or Soil-type Calibration" Technique 

Soil salinity (Ed or Ed) can be determined from the measurement of Ed, or Ed , made at 
approximately a reference soil water content using a calibration either established or predicted for 
the particular field or soil in question. Such calibrations are essentially applications of Equation 
[11]. Such linear calibration relations have been reported by Rhoades and Ingvalson (1971), 
Halvorson and Rhoades (1974), Rhoades (1976), Rhoades et al. (1977), Halvorson et al. (1977), 
Rhoades (1981), Rhoades and Corwin (1981), Cameron et al. (1981), Corwin and Rhoades 
(1982), Williams and Baker (1982), and by Slavich and Read (1983). These "pioneering" 
findings provided the impetus to the use of four-electrode and EM instruments to survey soil 
salinity that has occurred since then. Numerous satisfactory field calibrations (r > 0.9) have been 
obtained using this empirical technique for many areas, fields and soil types around the world and 
successfully used to diagnose and map soil salinity. Examples of such calibrations were given 
earlier (see Figures 19 and 20). Other examples are shown in Figures 48 and 49 for 



Soil salinity assessment 



49 



representative soils of the Northern 
Great Plains region of the United 
States. Various methods for 
establishing these types of calibrations 
are described in Annex 1 . 



FIGURE 48 




Relationship between soil electrical 


conductivity 


(EC a ), as determined with different interelectrode 


spacings and measured average soil 


salinity (ECe) 


for a glacial-till soil in Montana, 


USA (after 


Halvorson and Rhoades, 1977) 






c 


i i i i i i 
RICHEY SiCl 








<> 40 


Can-TOd , Manhma x 


an ff a 






SokfTMJ Form 








_~ 55 


*<i EC.-5HE<V9ifl / 


- 








r -C3H / 








5 5Q 
LJ V 


D A 








its 


□ y^ 








3 


a^n 








tS £0 


jfi a 


— 






O 


-^ IW#nlKf:rfl 


=.:■..!.-,■' 






f » 


/i ■ *~*' tm 


cm 








u 


^4 * » 


0- 30 






^ 10 


«/ a *° 


0-SO 






£ 


>f & *° 


0- so 






J 9 


■ v* ° ,w 


0-lW _ 






a 


/* i • i i ■ i 


1 1 






Jj L - 

c o i 2 3 * s e 


T B 3 




Electrical CondutliYify o! Soil, EC. 


dS*l 





Since water content affects soil 
electrical conductivity somewhat, as 
well as the relationships between Ed 
and soil salinity, determinations of Ed 
and the salinity-calibrations are made 
preferably when the soil is near field 
capacity. However, measurements and 
salinity appraisals can be made at 
lower water contents that exceed a 
certain minimum level, as discussed 
previously. For irrigated soils, 
measurements and calibrations ideally 
should be made after irrigation when 
the soil water content is at field 
capacity. This water content is 
sufficiently reproducible for such 
practical calibrations. Under dryland 
conditions, calibrations and measure- 
ments should be made in early spring, 
or on fallow land, in order to take 
advantage of the relative uniform 
conditions of soil water that exist then. 
In any case, these empirical 
calibrations should be established so as 
to apply to field soils with their natural 
structures, pore size distributions and 
water holding properties. Though 
Johnston (1994) has reported that he 
found calibrations established in the 
laboratory using disturbed samples 
were not different than those found 
under field conditions, this has not 
been the author's experience. If one 
desires to establish the kind of 
calibrations described in this section, it 
is recommended to do so under 
conditions that are as close as possible 
to the field conditions anticipated in 
their applications. This will maximize their accuracy and appropriateness. This is especially true 
if the EM-38 is to be used to measure Ed, since its depth- response function will vary with the 
distribution and magnitude of Ed in the profile, the presence of a shallow, saline water table, and 
certain other soil properties. 



FIGURE 49 

Relationships between soil electrical conductivity 
(ECa ) and salinity (expressed as ECe) for 
representative soil types of the northern Great 
Plains, USA (after Rhoades and Halvorson 1977) 





i j i i t i i i i i i 




4D - 
35 - 
SO - 
H - 
20 - 
IS - 

to - 


/ttO- . CL ,S*Cu X jT 

ff ■<</ s 

fur y 

/ / s*y 

// ' l x 

ff jf EC-it. e*fc**f* h T«M *-— 


? 


5 - 


// jf* Sjrt»W Tlilkrt |¥*Mfc mmm * i i ■■ n | 


,YS L hB* (iJHifii \ 
£f^ 1*CL ib'Jr iWj-t«»fi t** ****** - iDlnNn 




f 1 it :!:> | ::i T wt^atvi* 

lie imp on 


C 


r T 1 1 r l r l l i t 
) 1 2 3 * S G 7 9 ft 11* II 1 

Elitrrttnl Cwductitliy ** Soil . EC a .dS^m 



50 Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



Given the value of EC W at field capacity (or EC e ), one can readily estimate EC W at lower 
field water contents using Equation [8]; the values of EC W (or osmotic potential) occurring over 
the irrigation cycle can then be estimated and used to predict crop response to varying irrigation 
management. Of course, this estimation ignores the precipitation of salt that occurs with the 
reduction in water content. If more accurate estimates are required, corrections can be made for 
this latter process, as well as for some others, using computer models as discussed earlier. For 
many practical applications, the "compensation" phenomenon (implied in Equation [8] and 
discussed earlier) precludes the need to measure EG, per se, or to be too concerned about 
measurements of EG having to be made exactly at calibration water content. 

The measurements or bulk soil electrical conductivity may be made using a four-electrode 
sensor or an EM sensor. The only difference is that the latter measurement provides a more 
depth-weighted value of EG (i.e. EG ). If the soil is essentially uniform in EG and texture, both 
sensors will read the same, as shown or implied by the results of Rhoades and Ingvalson (1971) 
and Rhoades and Corwin (1981), and their salinity calibrations (i.e., Equation [11] relationships) 
will be the same. That is to say, one should obtain essentially the same linear relationship 
between soil salinity and either EG , or EG , irrespective of the type of sensor used for soils of 
uniform properties. However, one does not expect the same calibration to be obtained from the 
different sensors when EG varies substantially with depth in the soil profile. This is implied by 
the finding of Corwin and Rhoades (1984) which showed that the depth- weighting response of the 
EM-38 sensor was different for soils whose EG values increased with depth compared to those 
whose EG values decreased with depth. It was conclusively shown in the study of Rhoades et al. 
(1990b) that the EM-38 depth-response relation varies with the magnitude and distribution of EG 
within the 0-2 m depth of soil. This phenomenon has important implications with respect to 
deciding whether it is preferable to estimate soil salinity directly from the EM readings (i.e., to 
establish EG = f (EM) relationships) or to first estimate EG values from a sequence of EM 
readings and then to estimate salinity from EG (i.e., to establish EG = f (EG) = f (EM) 
relations). The advantages and disadvantages of the two approaches will now be discussed. 

Because calibrations vary with soil type and because many soils, especially alluvial arid 
land soils, have strata in their profiles which vary in texture and water content, it is obvious that 
no single EG (or EG 5) = f (EG, or EG ) relationship can be established for such a non-uniform 
soil which will apply to all of its different layers/strata/depth-increments. It is often important to 
know the salinity distribution that exists within the various depths and strata of a root zone; for a 
variety of obvious reasons, knowing the mean level is not sufficient. It was for these reasons that 
the author and his collaborators, from the very beginning in their use of the surface-array four- 
electrode and EM-38 sensors to determine soil salinity, decided to develop ways to estimate the 
EG values within different soil-depth intervals (from a succession of fixed-array four-electrode 
and EM-38 sensor readings) and, from these values and knowledge of the soil-textural properties 
within the various depths, to estimate (using texture-based calibrations) the salinity level for each 
significant region of the soil profile. The earlier means developed for predicting depth 
distributions of EG within soil profiles from a succession of EM-38 readings has already been 
described; analogous means using a sequence of surface-array four-electrode readings will be 
described later. Also described later are the newer methods that the author and collaborators, 
have developed for calibrating these sensors so as to able to predict soil salinity within various 
depth-increments of the soil profile. Of course, when the sensor measurements are made with an 
insertion EC-probe, one obtains directly the EG within each specific depth of interest (an 
advantage in accuracy that is associated with this sensor). Other users of the EM-38 sensor, 
especially those in Canada, have chosen to correlate the EM-38 readings with either mean profile 
salinity or the salinity weighted by depth in accordance with the uniform-soil depth-response 



Soil salinity assessment 51 



nature of the sensor (Cameron et al, 1981; Wollenhaupt et ah, 1986; McKenzie et ah, 1989; 
Slavich and Peterson, 1990). The author believes that they did this for three reasons: 1) the soils 
they work with are not highly stratified (they are mostly glacial-till soils, in the case of Canada), 
2) their primary objective has been to map general and gross conditions of salinity, and 3) they 
did not have a good method for estimating Ed by soil depth from their EM-38 readings. Johnston 
(1994) concluded that these latter approaches, as well as his method for estimating the mean 
profile salinity value from the mean (EMh + EMv ) reading, were more accurate and simpler to 
use, hence preferable, compared to approaches which first estimate Ed and, in turn, EC e . He 
based this conclusion mostly on the assumption that a one-step procedure would entail less error 
than a two-step one and on the results of a limited test made on a relatively small number of 
South African soils. This assumption/premise is not inherently valid. Nor is the test convincing, 
since it was based on estimates of average profile values. For the reasons given above, different 
calibrations are obviously required for the different textural-layers of stratified soils. Thus, the 
author recommends that, where it is useful to know the salinity distribution in the root zone, the 
surface array four-electrode or EM-38 sensor readings be converted to their depth-increment Ed 
values and these latter values be used to estimate salinity for the various important soil-depth 
increments of the root zone, or that direct regression relations be established between the sensor 
readings and the salinity levels for each important depth-increment/strata of the root zone (a 
stochastic method for obtaining the latter calibrations is given later). Of course, if only simple 
maps of gross spatial differences in salinity are needed for general characterization purposes, then 
the "mean" or "weighted" calibration approaches may be suitable and used. 

Information about salinity within discrete soil-depth intervals can be obtained by one of 
three methods: 1) measurements of ECa can be made directly within the desired depth-interval(s) 
using an insertion four-electrode probe (see Figure 37), 2) the Ed values within different depth 
intervals can be estimated from a sequence of variably-spaced surface-array four-electrode 
readings (using methods described below), or from variably configured EM-38 readings (using 
methods described earlier for the EM-38 sensor based on Equation [23], Figure 47 and Tables 1 
and 3) the Ed values within different depth intervals can be estimated from a sequence of 
variably-spaced surface-array four-electrode readings and/or variably configured EM-38 readings 
using directly established depth-specific sensor-calibrations. The latter stochastic method is 
described in a later section. Of course, more accurate results can be obtained from the direct 
measurements of ECi made within each depth-interval using an insertion EC -probe (PJioades and 
van Schilfgaarde, 1976). 

ECa values within various soil depth-intervals, hereafter designated by Ed , can be 
estimated from the sequence of ECa values obtained with a surface-array of electrodes and 
successively increasing current-electrode spacings using the following relation: 

ECai - (ai-l) = Ed = [ (ECa • & ) - (ECai-1 • fc-1 )]/(&- ft-l ) , [25] 

where a represents the depth of measurement and a-i represents the previous depth of 
measurement. The conventional use of this equation is based on the following assumptions: the 
depth to which conductivity is measured is equal to the one-third of the spacing between current- 
electrodes (or the space between each pair of electrodes when configured in the Wenner-array) 
and that the stack of soil electrical resistances of a sequence of "stacked" soil layers behave 
analogous to resistors in parallel (Barnes, 1954). Some good results have been obtained using this 
approximation (Halvorson and Rhoades, 1974; Rhoades and van Schilfgaarde, 1976), as shown 
in Figure 50. 



52 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



Alternatively, the EC a value 
within a particular depth-interval can 
also be estimated from a succession of 
EM-38 readings made at different 
heights and coil-orientations, as 
explained earlier. An example of such 
predictions made using the earlier 
procedures (Rhoades and Corwin, 
1981; Corwin and Rhoades, 1982) is 
shown in Figure 51. More generally 
accurate predictions can be made using 
Figure 47 and the relations given in 
Table 1, based on Equation [23]. 

"EC-Model" Technique 

In attempts to apply "generic" 
soil/field-type models of the kind 
described above to areas mapped as 
the same soil-type, it was found that 
the estimates of salinity were 
sometimes not sufficiently accurate 
because of the substantial variability in 
soil properties that existed within the 
mapping unit. In other words, areas 
depicted in soil survey maps as 
homogeneous soil-types were found to 
vary considerably in soil-type within 
the mapping- unit. It was decided that 
accurate estimates of salinity for such 
conditions (they are not unusual; fields 
can be quite variable in soil texture) 
would require an examination of the 
soil profile at each measurement site 
and the application of an appropriate 
calibration for each different type of 
soil encountered. The following 
technique for determining soil salinity 
from ECa was the outcome of this 
experience and decision. Essentially, it 
amounts to the practical application of 
the EC -model described in Chapter 3, 
in which field-estimates of the percent 
clay content of the soil and its percent 
water content relative to field-capacity 
(for each site and depth of interest) are 
used along with empirically determined 
relations to obtain the required model- 
parameters needed to solve Equation 
[5] for salinity, given knowledge 



FIGURE 50 

Relationship between EC X , as calculated from 
Equation [25], and soil salinity (expressed as ECe), 
for soil-depth intervals of 0-30, 30-60, 60-90, and 
90-120 cm for a glacial-till soil in Montana, USA 
(after Rhoades and Halvorson, 1977) 



Si 



(J 

III 



RkCHEY SiCl 

Conrad, Monlono 
Bolirtm Farm 



i 

o 
>■ 

> 

9 

TJ 

c 
Q 

I 



35 
» 



IE - 




O I /■ .1 *■ i ft ' b * i: 
Electrical Conductivity of Soil. EC,, dS/Wi 



FIGURE 51 

Graphs of measured and calculated (by three 

different methods) EC a -depth profiles for three 

California USA sites (after Corwin and Rhoades, 

1982) 

Elictrical Conductivity of Sort, EC,, dS/rn 
o s io is £0 



Q 15 



E G4 5 







Labniew ArW 
Sim I 

5 

HHlflCtord •QBWQVfr 
- h sidelined coeHirer- 

aawuch- luorttiJYi 

ffCFTfncnb- 
■d HtWfil witti taw 

pafai 

b EitnbJiitmcl czMfficunr- 

irwnrta from 
■ *,t ?.•■;. ir. 



% ^ 









Soil salinity assessment 53 



(measurement) of EG. These estimates of percent clay and relative water content are simply 
made in the field by "feel". These estimates are deemed sufficiently accurate for practical needs, 
as shown in Rhoades et al. (1989c, 1990a). 

Thus, in this approach, soil salinity is determined for each different condition of soil-type 
and water content, provided the latter is in excess of the threshold value, encountered at each 
survey location from the solutions of Equations [5] or [6], [7] and [8] (i.e., the bulk soil electrical 
conductivity model and the relation between EC W and EC e ) using measurement(s) of Ed, 
estimates of soil clay percentage and percent water content relative to field capacity and the 
following empirical relations to estimate pb , 0s , 0ws, 0fc and EG : 

SP = 0.76 (%C) + 27.25 , [26] 

p b = 1.73- 0.0067 SP, [27] 

6s = pb / 2.65 , [28] 

@fc = SP(p b /200) , [29] 

0. = 6fc (FC/100) , [30] 

0.,= 0.639 0, +0.011, [31] 

and 

EG = 0.019 SP- 0.434, [32] 

where %C is clay percentage as estimated by "feel" methods, tc is the volumetric water content 
at field capacity, and FC is the percent water content of the soil relative to that at field capacity 
as estimated by "feel". 

Given the above assumptions, estimates and measurement of EG, EC W is calculated from 
the solution of Equation [7]. Then EC e is determined from Equation [8], assuming that EGc = 
ECws and, therefore, that (EG, 0») = (EGc 0wc + EC,, WS ). These calculations can be made 
simple using a programmable pocket calculator; alternatively, EG can be obtained graphically 
using Figure 52, after Rhoades (1990b) and Rhoades and Miyamoto (1990). Examples of the 
successful use of this technique are given later. 

Sensitivity analyses and results of field tests have shown that the estimates and 
assumptions described above are generally adequate for practical salinity appraisal purposes of 
typical mineral, arid-land soils (Rhoades et al. 1989c and 1990a); i.e., that EG can be estimated 
in the field sufficiently accurately for most salinity appraisal purposes from the accurate 
measurement of EG and reasonable field estimates of %C and FC made by "feel". For organic 
soils, or soils of very different mineralogy or magnetic properties, these estimates may be 
inappropriate. For such soils, appropriate estimating procedures will have to be developed using 
analogous techniques to those used by Rhoades et al. (1989a). The accuracy requirements of 
these estimates may be evaluated using the relations given in Rhoades et al. (1989c). 



54 



Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



FIGURE 52 

Relationships between electrical conductivity of bulk soil (EC a ), electrical conductivity of 
saturated-paste extract (EC e ), relative soil water content as percent of field-capacity, and soil 
clay content (% clay), for representative arid-land soils (after Rhoades and Miyamoto, 1990) 



tf-E 



i .* 



lh^n4«l hriLC^BE^J m 






1 Mill II 




/ .' J , J ■ 


r- Uf.] 1,-iii p 


!■■»- ( ( '" ' ' - t* 


'' KJ--JL 4«*^n 


- -\- z • # * ; ffi 


■ 




-\ — i + -*&7r%- — 


: : : 


: 








j * *sp f 






. V <r 'J f 






•'j'sy> 


















iW-. - ■•■ £- . 


■ 


si 






' 










?4' 


±± , d 






:i: : ::::: :::: ^ 






:i:---::: 












- - ! 


XEl. 


. 


.IT - - -z--l 



an n.i 10 u iji 11 



mntnl rnUniHr pf HJt SM. BC.Jim 



.IkmrfWCw^MH «Sim 




nn cu 10 u 211 tl M *S 



'hniriill'.h ?.,-.,. «4rai 






: 11 ■ 11 : 1 . 11 11 ■ 


1 




































' 


j: ' 
























































J . ' 






















- 






























































. 




/" 












































































1. 






































































































■ 






























- 






ij j 





























I- 




















- 












- 




















- 


























































































































1 

























q^iU-C^dunt, dUiH.IC,U 



jyinMirtT^MC^in^JQaTM mmj 




'»uc»"ll V »"0» 





... j .,,,.... , 








Lftjjy 


?; 




«.-■:-.* ti«zs fe-IH 

3F- M *-, - JB 










ftt 


?'" 




















^Js 




































































































































— 






J 1— 


^3 




















* 








,-^ 








— t 














































1 


































































__ 








W-*~ 














=£= 






































1 . 1 













































































































































































Lk_Jr-~lUi-rJl.il-, I.' £1 _•_!.. tl. LX^.E-iTl 



I 
if. 



ft»« »f pma^i mail ■»_ 





1 . 1 . 1 . U 1 1 . 1 
















." r"" 




-• 








































, 
















; 


























■ 


















-Y 








- 


































































■ 














. 












" 


















































































>■> .- 






















- 






























































































■ 


















- 


- 






- 


















■ 














































■ 












- 




" 




































— 








4- 




7 


— — 












































: 


— 














































































- 




r- 


■ 




■ 


" 












































■< 


... 















EkonciJ CoKtaxmrj r/ It* Sal, CC V dEta 



JgWri<frMJDf*^iMaiQ 




tf Mli £UL «_,,-&■, 



i =; 























P-M-BtfH 


li 


■dryifWTUW 


























EP.it +,.* 






































































s 










































































L 












?i 














































































I 




































:> 


















































































































-4- 




:■! 


























■ 




























. 


-\- 


































-h 






































■■ 




























■■ 













































































































: ::: 


































- 





SinHCMiTtwXium iTtiui 



Soil salinity assessment 



55 



AiMm if fuM >>»ih/' SS« 




























ftanarfrUICjfM 


T m..ii 








- 












/-'• 


! 




















A 




.. 




Kb?.!! „.J(J p,t 
















' 


.: 
























:' 




■ 
















' 






























































































































































■ 


















































































































































































■Jv 




































































_l_ 




























■■] 










— 1— 








p*" 










































































































1 * 




































■ - 




















































































































L | 


























— 1 — ' 






. 












( 








1 


til 

■nH CllUll III- « IMS*', W,. B* 






'- 



























n_ jrMfl(Hn*]ij S(mtO 






j 






t 








#■ 




1 




■ae-H (h;- its it-ifi 














:■■" -" 




V 


■ » 


» 


> 


-■ 












' 








^^™ 








































































i 




£ it 
































i 


































t 


















































































































































1 
























f ■ 








































.„ 


— 












H- 




































— i 


i 


































j 


































j J 




































































5 





































































































! 4 * • 

BfeflM C«4w4<t) * ■* KU, 1,1111 

























■ IrMn- 


Pkia£*tf*r 


ft 


«m«v„ 






1 














^ - 


^ 


I... 




KZlj.lH fj.l» f,-l,)l 




















































































































































































7$& 












































- 










































































/Hi 
















E 


















_ _ 














^^^ 








9 ; J 


















hS 












"_ 


bb 








































: 










1 








































il 






























— | — 










1 1 






- 


























1 










i 




























i 














1 












1 
























■ 1 






— I— 




— 1 









1 LYlAa-l^ rfBlUM SC r JlW- 



If more accurate determinations of EC e , or EC» , are required than can be obtained by the 
estimation procedures described above, then quantitative measurements of @» , EC S , pb, etc. 
should be made using appropriate methods and used in place of the above-described estimates. 



The major advantage of the "EC-model/field-estimates" method described above is that it 
accounts for the site-to-site variabilities in soil properties (clay and water content in particular) 
that can occur within individual fields, mapping units, or other areas of interest. Essentially, it 
generates a specific calibration between Ed and EC e , or EC» , for the particular soil condition 
encountered at each site of Ed measurement in the field (area) under evaluation. Such "specific" 
calibrations are generally more accurate than are the "average soil-type" calibrations when 
applied to an area of field size or larger which is assumed to be the same as the calibration soil- 
type (Rhoades et al. 1990a). Field tests of this method have shown it to be sufficiently accurate 
for the practical purposes of salinity diagnosis and mapping, to be faster than conventional soil 
sampling and laboratory methods (measurement of Ed per se, either directly or as estimated 
from EC P ) and to be generally more accurate than the "soil- type" calibration (Rhoades et al. 
1990a; Annex 1). Another advantage of this approach is that it saves the time involved in 
establishing calibrations for each of the different kinds of soils found in the survey area. The 
major disadvantage of the method is that one must estimate (by feel) the clay percentage and 
relative water content of the soil at each site and for each depth of measurement. This requires 
that the soil be probed at every measurement site and that time be taken to make these estimates. 
Of course, when one collects sensor readings at any "unexamined "site and estimates the 
corresponding levels of salinity, especially within the various depth increments of the root zone, 
by one of the other approaches, there will always be uncertainty about the properties of the soil 
within the profile at that site and, hence, about the appropriateness/applicability/accuracy of the 



56 Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



predictions. As will be discussed in more detail later, the combination of the EC-model with the 
stochastic field-calibration approach, which is described next, is very appealing. 

Stochastic Field- Calibration Technique 

As explained earlier, when measurements of Ed, or Ed , are made using the mobilized and 
automated sensor systems, it is not possible to simultaneously estimate or measure the other 
secondary soil properties that are required to use the "EC -Model" technique described in the 
preceding section. For that reason, the "stochastic field-calibration" technique that was mentioned 
earlier was developed. This latter method is also applicable to hand-held sensors and is more 
rigorous than the "generic field or soil-type calibration" technique. This stochastic field- 
calibration technique is essentially a statistical/ground-truthing approach in which a predictive 
(regression) relationship between Ed, or Ed , and Ed (or Ed) is established and, 
subsequently, used to determine salinity from sensor measurements in a calibrated area of land 
that is relatively homogeneous with respect to soil conditions other than salinity. In this approach, 
spatial regression modeling techniques are usually preferable to the classical geostatistical 
modeling techniques, since the former require less calibration data and are typically easier to 
estimate (Lesch et al., 1995a, 1995b). 

The numerous sensor readings of Ed are obtained within the sampling area (usually a 
field) under evaluation on a uniform (centric systematic) grid basis. Based on the observed field 
pattern of Ed, or Ed , readings, a relatively small number of sensor measurement sites are 
chosen for soil sampling using a statistical model/procedure, which is described in more detail 
later. Soil samples are collected at these sites and their salinities are determined by any accepted 
method of salinity appraisal (the EC-paste method of Rhoades et al. 1989b is recommended for 
this purpose). A multiple linear regression relation is then established between the Ed and/or 
EC, readings, the measured soil salinities (Ed, Ed, or some other expression of salinity), and 
the x/y coordinates for each soil depth of interest. Such a spatial regression model can be written 
in matrix notation as: 

F=X(3 1 +W(3 2 +8, [33] 

where Y represents the vector of log transformed soil salinity values, X represents a matrix of log 
transformed and de-correlated sensor readings, W represents a trend surface matrix based on the 
spatial coordinates of the measurement sites, and e represents a random error component. For the 
fairly typical case where W is a 1 st order trend surface matrix, equation [33] becomes: 

\og(EC e ) = |3 + pj \og(EM H )+ $ 2 [\og(EM H )- log(i?M y )]+ P 3 log^C 4/ J + P 4 X + PgF +e [34] 

where EMv and EMh are as previously defined, Ed P refers to Ed as determined from the four- 
electrode sensors, (X,Y) refer to the spatial-coordinates of the measurement sites, [3 represents the 
regression-fitted parameter estimates, and e is the random error component. The resulting field- 
specific relation is subsequently used to predict the salinities at the vast number of unsampled 
sites/depths in the area where the remainder of the EM-38 and four-electrode sensor 
measurements were made. This "single-step" method eliminates the need to first convert EM-38 
(essentially Ed ) readings to equivalent Ed values within a particular soil depth using general 
relations (such as those in Table 1 which are based on Equation [23]) and thence to Ed , as is 
required by the "EC -Model" technique. 



Soil salinity assessment 57 



Software (Estimated Salinity Assessment Programme, ESAP) has been developed to 
determine the regression model and to prepare a map of the predicted salinity pattern for a 
surveyed field; the procedures for use of this software are described, with examples, in a User 
Manual (Lesch et al., 1995c). The ESAP software also provides an algorithm to determine the 
numbers and locations of sites in a surveyed field to be soil sampled for calibration purposes; this 
is discussed more in a following section. This software is available from the US Salinity 
Laboratory. A "windows-based" version with various "user-friendly" features is now under 
development; it should be available by the time this paper is published. 

Experimental results show that this method works very well for fields/landscapes that are 
relatively homogeneous in all factors affecting EC a conditions other than salinity, such as 
individual fields uniformly managed or sections of natural landscapes that have similar soil types 
and properties (certain dryland landscapes for example). This approach substitutes easily 
acquired EC-sensor field measurements for the more difficulty carried out procedures of soil 
sampling and laboratory analysis. It very substantially reduces the number of soil samples 
required by traditional methods to accurately and intensively map the spatial salinity patterns 
within fields, as well as the overall cost. Larger areas of land can be mapped by joining adjacent 
areas on a field by field basis. This method is more practical than those based on conventional 
geostatistical procedures, such as those traditionally used for salinity mapping purposes (i.e., 
Webster, 1985, 1989), because it reduces the intensive soil sampling generally needed to obtain 
the accurate variogram estimates required in these latter procedures (unpublished data). The 
major limitation of the method is the requirement that the fields be under relatively uniform in 
management and that soil water, bulk density, and clay content be reasonably homogeneous. If 
needed however, larger fields (or areas) can be subdivided into smaller more homogeneous units 
and the method applied analogously to each sufficiently homogeneous subunit. Alternatively, 
additional practical measurements besides ECa, such as location coordinates, elevation, etc., can 
be made and incorporated into the regression relation (such as coefficients p% and (3s do in 
Equation [34]) to adjust for some of the "other" factors influencing the salinity prediction (Lesch, 
et al. 1992). Disadvantages of the method are the need to enter the field a second time after the 
ECa -sensor readings have been taken to locate the selected sample-sites and to acquire the 
"calibrating" soil samples. The latter locations are not difficult to establish when numbered 
markers have been left in the field at the sites of each Ed measurement, or with the use of "real- 
time" GPS systems. This need for re-entry is not a major factor when large areas are being 
mapped. A soil sampling team is usually sequenced one-day after the ECa measurement 
operation; the statistical calculations used to select the sampling sites can be made at field-side by 
another team member. This method is especially appropriate where very rapid, mobile 
instrumental systems are being used to intensively map large fields or areas of land. 

While the use of both EM-38 and four-electrode measurements are included in Equation 
[34], analogous relations can be developed using just EM data or just four-electrode data. 
Furthermore, the relations can be developed for a single soil depth or for a series of soil-depth 
intervals. These empirical relations typically yield highly accurate predictions of the spatial 
pattern of soil salinity in fields, since the model is specifically calibrated for each field (Lesch et 
al., 1992, 1995a,b). They are highly field-specific and can not be developed without calibration 
data. Such relations can also be used for monitoring purposes (i.e., testing for a change in the 
condition of salinity over time), provided additional soil samples are acquired at the future time 
(Lesch et al., 1998). This use is discussed later in the context of salinity monitoring. 

Since the major effort involved in this "stochastic-calibration" approach stems from the 
collection and analysis of the soil samples, one tries to minimize the number of calibration sites. 



58 Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



The location of these sites are chosen so they meet certain statistical criteria, such as the optimal 
estimation of the regression parameters and /or the minimization of the prediction error. An 
algorithm was developed for generating such a model-based sampling/calibration methodology 
for use with the mobilized combination-sensor, salinity assessment systems described above (but 
which has more general applicability) and it is described in Lesch et al. (1995a, 1995b) and in the 
Estimated Salinity Assessment Programme (ESAP) Software (Lesch et al., 1995c). 

The good success obtained in determining the levels and spatial patterns of soil salinity in 
irrigated fields and crop root zones using the various above-described methods may be seen in the 
following publications (Lesch et al., 1992, 1995b; Rhoades, 1992b, 1993, 1994, 1996b; Rhoades 
et al., 1997a, 1997b). Examples are given later which illustrate the typical success achieved with 
this stochastic field-calibration approach to measure and map salinity. While the above described 
approach has been found to be generally quite robust and accurate, it is not suitable if the field is 
very heterogeneous in soil type. For such situations, the EC-model (Equation [5]) method is 
advised, or the combined use of it with the stochastic-calibration method. 

Comparisons of the Different Methods of Measuring Soil Salinity 

Only a few direct comparisons of the various instrumental and conventional methods of 
measuring soil salinity have been made to date. Salinity measurements were made by four 
methods in a field experiment in India (Yadav et al.,\919), i.e., porous-matrix salinity sensor 
(ECm), vacuum-cup soil water sampler (EC), soil samples (EC e ), and four-electrode soil 
conductivity sensor (EC; surface Wenner- array method). These investigators found a better 
linear correlation between EC and EC (r = 0.93) than between EC and EC (r = 0.78) or 
between EC and EC (r = 0.78). They concluded that, for purposes of diagnosing the salinities of 
the soils of an extensive area, the four-electrode technique is preferred because it is more rapid, 
simpler, and more practical. Loveday (1980) compared the four-electrode (surface Wenner-array) 
technique with soil sample extracts (EC) in a survey of 50 field sites in Australia. The water 
contents of the soils at the time of measurement were not generally at field capacity. He obtained 
relatively high correlations between EC and EC , though variance was high. He attributed this 
high variance to field variability factors and concluded that the four-electrode method was good 
for gross survey work but not accurate enough for predictive purposes. However, Loveday used 
generic soil-type calibrations and only two 5 cm-diameter soil samples to estimate the salinity of 
the relatively large volume of soil included in the Wenner measurement. One must question that 
such small samples represent "ground truth", and hence the appropriateness of his conclusion. 
Indeed, it has been found that the so-called salinity "ground-truth", as typically determined 
using small-volume soil samples, used to test the credibility of instrumental techniques of 
salinity appraisal are usually not very representative of the larger volume of soil involved in 
the instrumental measurements (PJioades et al. 1989d, 1990a; Lesch et al. 1992). Loveday also 
concluded from his results that EC - EC calibrations found in the US by the author and 
collaborators were probably universally applicable to soils of similar texture. Van Hoorn (1980) 
compared salinities measured using extracts of soil samples with both those obtained by four- 
electrode surface-array and four-electrode EC-probe methods in large experimental tanks. He 
concluded that for survey work either the Wenner method or the four-electrode probe could be 
used, but that the accuracy of the latter is much greater. Nadler and Dasberg (1980) compared 
soil salinity measurements made in small salinized field plots using in situ ceramic porous matrix 
sensors, four-electrode EC-probes, a four-electrode Wenner-array, and soil sample extracts (1:1). 
They found good correspondence between "expected" salinity and both "soil extract" salinity and 
"four-electrode" salinity, but not with "porous matrix salinity sensor" salinity. They attributed the 
latter discrepancy to lag-time problems. They concluded that the Wenner-array method could be 



Soil salinity assessment 59 



used more reliably under drier soil conditions than could the four-electrode probe, which requires 
better electrode-soil contact for accurate measurements. Johnston (1994) evaluated the 
suitabilities of four-electrode and EM-38 sensors for appraising soil salinity, as well as various 
means of their calibration. The authors has already discussed most of his findings and 
conclusions in previous sections of this paper. He found that the methods were suitable for 
practical salinity diagnosis and mapping purposes, given appropriate site-specific calibrations, 
and concluded that the EM-38 was more convenient and accurate. As the author has already 
stated, he does not agree with all of his conclusions for the various reasons previously given. 
While he made the most thorough and field-based evaluations of the attributes of the various 
sensor-methodology that is the subject of this paper, much of it was directed to the earlier less- 
developed procedures than those advocated herein. 

The author obtained good results with the use of both four-electrode and EM-38 sensors 
and with all of the various methods of soil salinity assessment described in this report, not only in 
the numerous locations in the US, but also in other countries. He has found these sensors and 
techniques to be useful in varied applications including salinity diagnosis, mapping, monitoring, 
saline seep and water table encroachment identification, irrigation scheduling and control, 
leaching fraction assessment, identifying areal sources of salt-loading, evaluating drainage 
adequacy, evaluating irrigation-infiltration uniformity, and in developing site-specific farming 
plans. The author finds the porous matrix salinity sensors to be less generally useful than the 
sensors which measure soil electrical conductivity, because of their small sampling volume and of 
their substantial lag time response to changing soil salinity situations. However, for some 
applications they may still be the preferred technique. He also finds the TDR sensors to be less 
useful for salinity appraisal than the four-electrode and EM-38 sensors, because the former are 
less robust, more limited in their volume of measurement, less adaptable to mobilization, more 
limited in the range of salinity they will sense, and more time-consuming in data acquisition and 
interpretation. But in fact, the author recommends that the various techniques be used 
complementarily; the mobilized and combined EM/four-electrode system is most suitable for 
surveying large fields in detail to establish the larger scale spatial variability in salinity conditions 
and the underlying causes of it, and the four-electrode probe is more suited to acquiring detailed 
information of EC a (and salinity) within various regions of the root zone, such as below the 
furrow, within the bed, with distance from drip emitters, etc. The fewest appropriate number of 
soil samples can then be taken from the different areas for detailed chemical analysis of the 
salinity composition, if desired, using the salinity variability information obtained with the 
instrumental readings and the ESAP software. This "combined-use" approach greatly facilitates 
the tedious, time-consuming and costly aspects of soil sampling. Whether the soil samples are 
reacted with water, or soil water per se is isolated from the soil for detailed analysis, is a matter 
of need and practicality. For practical reasons, aqueous extracts are generally used, although 
ideally one would prefer an analysis of the actual soil water. When an extract is to be used, it 
should be the one with the lowest water:soil ratio feasible. The EC-paste method has proven to be 
very accurate and dependable; it can be used advantageously in lieu of extracts when soil samples 
need to be analyzed only for salinity. The new SAR-paste method, referred to earlier, can be 
analogously used to diagnose and screen soil samples that need to be appraised, especially in the 
field, for sodicity problems (Rhoades et al., 1997c). 

The equipment and procedures described and advocated herein are all undoubtedly useful 
for many purposes of salinity assessment and can be used in many different ways; the most 
appropriate will vary with the exact needs and circumstances of the user. It is for this reason that 
the approach taken in this report is to provide a fundamental understanding of soil electrical 
conductivity, of the attributes of the various sensors and methods of their calibration and 



60 Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



interpretation, so that the various kinds of users can adapt them as needed to meet their specific 
applications and circumstances. Such users should always be aware of the limitations inherent in 
each of the alternative methods of measuring soil salinity and take them into account. The most 
appropriate one(s) should be used according to the specific needs and objectives of each 
particular situation. Again, the overall task of measurement and monitoring of soil salinity can be 
greatly facilitated through the combined use of the various methods. The EM and four-electrode 
instrumental methods should be used for most of the field characterization needs; laboratory 
analyses can then be carried out on only the minimum appropriate number of soil samples 
collected in accordance with the findings of the surveys made with the field-instruments. The 
areas requiring separate sampling are most easily determined from mobilized EM/four-electrode 
measurements; the depths to be sampled and numbers of samples to be taken from within each 
sampling area/depth are most accurately determined using the ESAP software. The most accurate 
salinity profile information can be determined from four-electrode probe readings and the EC- 
model relations. 

A comparison of the costs and time involved in making salinity assessments at the field and 
regional scales using the different methods and equipment reviewed above is given in Chapter 5. 
The results are summarized in Table 26, along with a comparison of the differences in the amount 
of information/data provided by the different methods. Only the overall results of this evaluation 
will be summarized here. 

The evaluation should be understood to be only relative, since the actual costs (both capital 
and operational) will vary from one country to another depending on differences in their labor 
costs and technical development. The costs are based on conservative estimates and US 
conditions; if there is bias in this evaluation, it is made intentionally in the favor of conventional 
soil-sampling and laboratory analysis methodology, so as not to over-promote the instrumental 
methods. The cost to undertake detailed field-scale surveys (for a typical 64-hectare field, 12 by 
12 grid and three soil-depths) using soil samples and traditional laboratory extraction procedures 
is concluded to be completely cost prohibitive ($146 per ha.; 43 hours of field time). The cost can 
be reduced to $26 per ha. using the EC-paste method of PJioades et al. (1989b), but with no 
savings in field time. The cost can be reduced even more with the use of field instrumentation 
which measures bulk soil electrical conductivity. Detailed field-scale surveys can be made using 
three different hand-held, instrumental approaches (two of which require the use of some soil 
samples for calibration purposes) at a cost savings of 96% compared to conventional method. 
Such surveys can be accomplished: 1) using the EM-38 at a cost of $6.50 per ha., requiring 4.4 
hours of field time, 2) using a four-electrode surface-array sensor (two soil depths only) at a cost 
of $7 per ha., requiring 6.2 hours of field time, and 3) using the EC-probe (without the analysis 
of any soil samples) at a cost of $6.60 per ha., requiring 14.6 hours of field time. With the use of 
the mobilized instrumental systems, the costs can be lowered even more when regional 
assessments are undertaken ($3.24-$3.68 per ha.) and are as cost effective as the hand-held 
instrument methods for detailed field assessments while providing substantially more spatial 
information in the latter case. An added benefit of the mobilized instrumental methods, compared 
to conventional methods and hand-held instrumental methods, is the very substantial reduction in 
the field time that is required by the former methods (the required time with these methods is 2.7- 
3.7 hours per 64-hec tare-field). It is concluded that the instrumental methods of soil salinity 
assessment described herein are very cost effective; they are also very time effective. As shown in 
this evaluation, the savings in field time is substantial with the use of the field instrumental 
methods. Another time savings feature of these methods, but not included in the 
evaluation/comparison, is the timeliness of the information/results they provide. With the 



Soil salinity assessment 61 



mobilized systems, a detailed field-scale survey can be completed in the same day. With 
conventional methods, weeks to months are usually required in this regard. 

Determination of Locations of Measurement and Calibration Sites 

For purposes of salinity mapping and many other salinity assessment applications, the sites where 
EC-sensor readings or soil samples are taken must be associated with geographic location (x- and 
y-coordinates). If rapid methods of salinity measurement are to be used to best advantage, then an 
equally rapid means of determining sample site location must also be used. For this purpose, the 
LORAN and GPS systems used in marine and aviation navigation can be employed, where it is 
available, with success for certain types of salinity mapping (see Rhoades et al. 1990a, 1990c). 
The LORAN-C system, which broadcasts pulsed radio signals at a frequency of 100 kHz, is 
operated by the U S Coast Guard in cooperation with several other countries as an aid to marine 
navigation. The coverage is very good regardless of terrain and is not limited to line-of-sight 
transmission because of the use of the LF radio waves. The LF radio band is propagated by 
means of the Ground Wave, so that the radio waves closely follow the surface of the earth. The 
receiver calculates its' location by measuring the time delays of the received signals from three 
different transmitter locations and applying the principle of triangulation. Thus, the LORAN-C 
receiver is essentially a precise time-difference measuring instrument which processes the 
received information to determine a position-fix. The position-fix is given directly in terms of 
latitude and longitude coordinates expressed in degrees, minutes and seconds. With local 
calibration the repeatability of position determination can be as good as 10 meters, or better 
(Rhoades et al. 1990c). This "accuracy" is good enough for regional surveys, but not so for more 
detailed mapping and small-scale assessment applications. 

The global positioning system (GPS) provides a more accurate and generally available 
means to establish sample-site positions for assessment and mapping purposes. The GPS is a 
satellite-based radio navigation system operated by the U S Department of Defense. It consists of 
21 satellites in circular orbit at a 20,000 - km altitude and provides world-wide, 24-hour 
coverage. The system is analogous in concept to LORAN, but utilizes the line-of-sight reception 
of signals from multiple satellite-based transmitters of known position. A GPS receiver unit 
obtains/deciphers coded and synchronized signals emitted by several (usually at least 3) GPS 
satellite - transmitters in terms of time of measurement, distance from the transmitter, and 
position of the receiving antenna. Distances are determined by measuring the difference in time it 
takes the radio signal to travel between the various satellites and receiver by means of accurate, 
synchronized clocks contained within both the transmitters and receivers. The receiver (sample) 
position is calculated by "triangulating" the range distance from three or more satellites of known 
position. These calculations are carried out by the GPS receiver. The Global Positioning System 
includes five control stations evenly spaced around the earth near the equator. They track each 
satellite, determine their exact positions and transmit correction factors to the satellites and, in 
turn, back to the receivers. The suitability of the simplest/cheapest GPS for soil survey purposes 
has been shown by Long et al. (1991) to meet the accuracy requirements for detailed soil surveys 
(30.5-m). Accuracy is increased by averaging multiple readings taken over 10 seconds and by 
post-processing the data obtained by the mobile-receiver data to correct it for "drift" using 
analogous data collected over the same time period with a fixed-base, reference receiver (base 
station). Accuracy's of receiver position to within 2 to 5 m of true is made possible through use 
of this so-called "differential-mode" of operation (two receivers; one mobile and one stationary) 
and post-processing technique. Positional accuracy, under the mountainous and forested 
conditions of western Montana (USA), was found to average between 3 and 4 meters in the open 
and between 5 and 6 meters under closed forest canopies (Gerlach and Jasumback, 1989). Real 



62 Determination of soil salinity from soil-paste and bulk soil electrical conductivity 



time, differentially-corrected readings can be obtained using radio receivers and transmitted 
"corrections" from either dedicated stations or your own base station. The latter procedure is 
used in my mobile salinity assessment systems to establish the coordinates of EC a measurement 
sites and of soil sampling locations (Rhoades, 1992a,1992b; Rhoades et al., 1997d). The 
accuracy of this system is about 20 cm. The differential corrections can also be made in real time 
by incorporating a portable PC/software into the GPS system. There are numerous companies 
now selling this kind of equipment; the reliability and convenience features have been increasing 
steadily as the cost has been steadily declining (some costs are given in Chapter 5). The 
technology is well developed and extremely useful, in fact almost a necessity for salinity 
assessment techniques based on the mobilized system advocated herein. 

In an earlier section, the stochastic field-calibration technique for predicting soil salinity 
from EC a -sensor readings was described and advocated as an accurate and efficient means for 
assessing salinity at the field scale. In this method a spatial multiple linear regression model is 
established for the surveyed field based on intensive sensor readings and limited soil samples 
collected and analysed for salinity. The intent is to minimize the number of soil samples used in 
the calibration while assuring that the calibration is representative of the whole field. A site 
selection algorithm has been designed to facilitate this calibration, and also to select sites for 
follow-up monitoring evaluations. The algorithm selects sites that are spatially representative of 
the entire survey-area and simultaneously facilitates the accurate determination of the model 
parameters, based on rigorous geostatistical procedures. The advantage of the algorithm is that it 
is more cost-effective compared to conventional cokriging methods; regression models can be 
fitted with substantially fewer calibration sample sizes than is required with cokriging. With this 
algorithm, a suitable stochastic field-calibration is obtained with a small number of calibration 
sites (n ~ 8-20, depending upon the accuracy requirements of the survey) by combining the 
survey site information with response surface design techniques. It ensures that the selected set of 
calibration sites (i) is spatially representative of the entire survey-area and (ii) is suitable to 
permit the efficient determination of the regression equation parameters of Equation [34]. 

This sampling-location algorithm is provided in the ESAP software package. The details of 
the procedure are described in Lesch et al., (1995a and 1995b). Briefly, this algorithm transforms 
and decorrelates the Ed readings by a principal components analysis; it uses the transformation 
in conjunction with a response surface design to identify a statistically efficient set of calibration 
sites, and it modifies the response surface design as needed to optimize the spatial locations of the 
final calibration sites. An earlier version of this approach is described and illustrated in Lesch et 
al. (1992). An example illustrating the locations of the sensor readings and the calibration sites is 
given in Figure 58, which is discussed later. 



Soil salinity assessment 63 



Chapter 4 

Example uses of salinity assessment 

technology 



The following examples are intended to illustrate the utility of the instrumental salinity 
assessment technology described earlier. It is not intended to show how to apply the equipment to 
each and every kind of problem and situation; there are too many combinations in this regard. 
Many of the examples are based on earlier studies carried out before subsequent improvements 
were made in equipment, spatial statistics, software, theory/models and empirical relationships. 
Thus, better results than those obtained in some of these examples could most likely be obtained 
today using the newer improved versions of the technology, especially the mobilized systems, the 
EC-model technique and the stochastic field-calibration technique. Still, the simpler earlier 
methodology can be useful for many applications. Thus, the examples given should be viewed as 
instructional and as illustrative of the many ways and opportunities that exist to utilize the 
measurement and assessment technology and as a means to help the readers envision other 
possibilities and manners of utilization; the possibilities are numerous. 



Diagnosis of Soil Salinity and Saline Seeps 

A saline seep is an area of formerly productive non-irrigated soil that has become too wet and 
saline for economical crop production, as a result of the flow of saline subsurface water to the 
soil surface. Seeps generally develop on the lower positions of hillsides where there is a change in 
slope. Typically, water percolates through the soil profile located in an upslope recharge area, 
picks up salt in the process, is intercepted by a slowly permeable horizontal stratum, moves 
laterally through the relatively more permeable layer, resurfaces where the permeable stratum is 
truncated on a hillside, evaporates and deposits the accumulated salt, forming a saline seep. The 
recharge area is that area upslope from the seep (discharge area) from where the percolating 
water originates. Typically, excess percolation is caused by the conversion of permanent 
vegetation of higher net evapotranspiration to an annual crop of lower usage. Problems in 
combating saline-seep development are associated with diagnosing soil salinity, identifying 
potential saline-seep areas, determining increasing soil salinity trends in the field, and detecting 
the encroachment of a shallow perched water table before excessive crop damage occurs. 

The levels and distributions of salinity in the landscape, especially in the soil profile, were 
shown to be good indicators of encroaching seep development, as well as of the likelihood of 
incipient crop failure. Plots of EC a , as determined by surface-array, four-electrode measurements, 
versus interelectrode spacings (essentially equivalent to soil-depth in the Wenner-array) yielded 
distinctively shaped curves for recharge areas, encroaching shallow water tables, and seep areas, 
as shown in Figure 53 (after Halvorson and Rhoades, 1974). The curves for seeps showed a 



64 



Example uses of salinity assessment technology 



sharp decline in EC a with depth 
indicative of salts being carried upward 
to the soil surface from the shallow 
water table and their accumulation at the 
surface by evaporation. In contrast, 
curves for recharge areas showed a 
gradual increase in EC a with depth and 
then a steady level with further depth 
corresponding to the leaching of salt in 
the soil profile and a net downward flux 
of water. Curves in areas that were not 
yet excessively salinized but which had 
water tables approaching critical depths 
displayed a level of EC a (salinity) that 
was significantly higher than typical of 
the normal regional soil. These areas 
also displayed an initial sharp increase in 
EC a with depth (to about 0.5 m) and 
then a gradual decrease with further 
depth corresponding to the interaction of 
leaching of salts in the near-surface soil 
and the upward flow of saline shallow 
groundwater into the soil profile. These 
salinity levels and distributions are 
distinctive and readily interpretable in 
terms of the processes of leaching and/or 
drainage. 

The presence of a water table 
within the critical depth (capable of 
contributing salts to the root zone by 
capillary flow) in the soils of the 
Northern Great Plains (USA) where 
saline seeps occur was found to be 
detectable from the level of salinity (or 
EC a ) in the topsoil (0-30 cm depth), as is 
illustrated in Figure 54 and in Table 2. 

The location of the recharge area 
of a particular saline seep was shown to 
be identifiable by mapping/tracing the 
subsurface pattern of high EC a levels 
upslope from the seep. 

An example is given in Figure 55, 
after Halvorson and Rhoades (1976). 



FIGURE 53 


Relationship between electrical conductivity of 


bulk soil (EC a ) and "Wenner-array" interelectrode 


spacing (approximately equivalent to soil depth) 


for a saline seep, an encroaching saline seep 


site, and an unaffected site for glacial-till soil in 


Montana, USA (after Rhoades and Halvorson, 1977) 




GLACIAL TILL SOIL 

Sidney , Mwitona 




3.5 


- Tvei( farm 

F Deplhl&Waler- 




® 3.0 


Silg Type toblt , metres 




J 


■ Saline 5e«p O.&l 




Ul 

t Z 5 


\ A SoliMSMp '■ 0T 
" * o Uftclftdtd £.SO 






\ 




-f Z.O 


\ 




■^ 


\ 




<3 1.3 


- ^V 




"5 






4 1.0 

LJ 








3.5 


1 r 1 t 




£ 4 6 ft 10 


Inter electrode Spacing , metres 



FIGURE 54 

Relationship between electrical conductivity of 
bulk soil (EC a ) in the 0-30 cm. depth- increment 
and depth to water table in typical glacial-till 
soils of Montana, USA (after Halvorson and 
Rhoades, 1974) 
1 



$ 

L 
ft 




RKtncat Gmduciwity. of Soil, ECo. dSAn 



Soil salinity assessment 



65 



TABLE 2 

Diagnostic EC a values for distinguishing unaffected soil sites, incipient saline-seeps and 

developed saline-seeps for representative soil types of the Northern Great Plains, USA 



Soil Type 1 


Site Condition 


Unaffected 


Incipient Seep 


Developed Seep 


EC a , dS/m 


C, SiC 

SiCI, CI, SCI, L 

SI 


<0.5 
<0.3 
<0.2 


0.8 
0.5 
0.4 


>2.5 
>1.5 
>1.0 



C = clay; SiC = silty clay; SiCI = silty clay; CI = clay loam; SCI = sandy clay loam; L = loam; SI 
sandy loam 



FIGURE 55 

Maps showing:(A) surface topography and location of saline seep, and marginally and 
unaffected alfalfa crop surrounding it; (B) isolines of EC, in the 0-30 cm. soil depth-interval, and 
(C) isolines of EC X in the 30-60 cm. soil depth-interval (after Halvorson and Rhoades, 1976) 






ECg IidHAi. dSfci 



ECl Iwllnii. dMn 




These figures show that the flow of water into the seep is from the north (there were 
several possibilities) and outward primarily toward the southwest. This methodology permitted 
the location of the recharge area to be identified and planted with permanent vegetation to help 
mitigate the problem. 

This example illustrates how the instrumental field-salinity assessment methodology may 
be used to facilitate the collection of spatial information about the levels and distributions of soil 
salinity in dryland soils, which in turn leads to the useful determination of the eminence of a 
saline seep problem and to the interpretation of the sources and causes of salinization in an 
affected seep-area and, thus, to meaningful management implications. Of course, the 
measurement technology also permitted the problem of salinity to be diagnosed and the areal 
extent of the salinized soil in the surveyed area to be mapped. This latter topic is the subject of 
the following section. 



While the example given in this section was based on the use of four-electrode 
methodology, EM methods could have also been used; however, when this work was undertaken, 



66 



Example uses of salinity assessment technology 



such equipment was not yet available. Likewise, the mobilized-sensor equipment and interpretive 
methodologies developed since then would also permit faster and more detailed faster surveys to 
be made now days. 



Inventorying Soil Salinity 



FIGURE 56 

Antenna, battery pack and meter for 

measuring location on the landscape using 

the LORAN technique (after Rhoades et al., 

1990c) 



Both near surface and subsurface salinity have been successfully mapped using several of the 
instrumental approaches/methods of salinity assessment previously described by collecting the 
instrumental measurements in relationship to spatial location and by displaying the data in terms 
of maps, or as "transect" plots, of EC a , EC W , or EC e . 



Example maps of soil salinity in the 
vicinity of a saline seep produced by the 
"soil-type" technique were shown in Figure 
55. A detailed evaluation of the suitability of 
the various sensors for measuring EC a and of 
the "EC -Model" technique (see Chapter 3, 
section Procedures for measuring bulk soil 
electrical conductivity) for converting EC a to 
EC e and for mapping soil salinity was 
undertaken in a 15-square mile (39-square 
km) sized study-area in California (Rhoades 
et al. 1990a). In this project, the instrumental 
measurements were made manually at pre- 
determined sites which were located by use of 
LORAN navigation techniques (see Figure 
56; GPS equipment was not yet readily 
available nor affordable at the time); the area 
was traversed on foot (practical mobilizing 
equipment had not yet been developed). 
Contour maps were made of measured and 
predicted salinities using data collected at 
about one thousand locations. EC e was 
predicted from EC a as measured by both 
four-electrode (Wenner-array and insertion 
EC-probe) and EM-38 sensors. The values of 
EC e (both measured and predicted) were 
plotted using SURFER (1986) software at 
both 200-m and 400-m grid spacings. This 




resulted in 1000 and 273 equally spaced grid nodes, respectively. Only contoured maps for the 
400-m grid spacing are shown here (see Figures 57A and 57B), because the 200-m grid spacing 
resulted in too many contours to clearly represent in maps of such scale. The corresponding 
values of measured and predicted EC e at each of these nodes were determined with the SURFER 
software. The proportions of the surveyed area by classes of soil salinity are given in Table 3. 
Such data can be used advantageously to assess the magnitudes of crop yield losses and 
associated economic losses caused by soil salinity. 



Soil salinity assessment 



67 



FIGURE 57 

Contour map of :(A) measured and (B) predicted soil salinities, in 15-square mile study area in 

Central California USA (after Rhoades etai, 1990a) 



Mi**»rt4 Soil SriiniritS. EC ( in rfS/ifl ; 40Onii[regrid basis 





I 



_s 



^1 I 1 



A mm 



^L 



1/7 \ A rr, 



y "^feri o YJ/A 



o 



©u 







PraJicled soilsslirrtitSvEC^indS/m: A<X-me\rt arid iHsla 




Visual comparisons of these measured and predicted 
salinity maps showed them to be essentially the same, 
irrespective of which of the three sensors was used to 
measure EC a . The absolute levels of salinity estimated from 
the three sensors were also similar (see Rhoades et al. 
1990a). Where the differences were substantial, the salinity 
levels were so high as to make the errors in estimate 
agriculturally unimportant. Of the three methods used to 
measure EC a , the four-electrode probe was found to be the 
most accurate, followed by the four-electrode surface-array 
and then the EM-38. The accuracy's essentially followed 
the degree to which the volume of soil measured by the 
sensor compared with that of the soil sample used to 



TABLE 3 

Salinity distribution in Kings 

River Watershed Survey Area 



Quantiles 


Percentage, % 


Soil Salinity; 




EC e , dS/m 


100 


79.8 (max) 


90 


15.6 


75 


8.1 (Q3) 


50 


3.5 (med) 


25 


1.2 (Q1) 


10 


0.7 





0.3 (min) 



68 



Example uses of salinity assessment technology 



FIGURE 58 

Maps of measurement and sampling locations and of 
measured and predicted soil salinity patterns in a field 
(Hanford S2A) located in the San Joaquin Valley of 
California USA (after Rhoades etal., 1997d) 



(a) 
38(1 

2SS 



S 190 

s 



95 


(b) 



Hanford S2A: Survey Grid 



90 180 270 360 
X, metres 



■ survey and calibration site 
• survey site 

□ ECe < 2 dS/m 

[=] 2 < ECe < 4 dS/m 

H 4 < ECe < 8 dS/m 

H ECe > 8 dS/m 



(<0 



Hanford S2A: Predicted ECe 



Hanford S2A: Observed ECe 



determine salinity (which was 
determined by conventional 
laboratory methods and assumed 
to represent "truth"). The 
accuracy of any of these 
instruments was adequate for 
practical salinity mapping 
purposes. Sample variability due 
to size differences in the volumes 
of soil used to measure salinity, 
which was small compared to that 
measured with the four-electrode 
surface-array and EM-38 sensors, 
was concluded to sometimes be 
appreciable and, when so, to 
result in an underestimate of the 
accuracy of the EC -model method 
of salinity appraisal using these 
larger-volume sensors under the 
surveyed field conditions 
(Rhoades et al. 1990a). 

These latter findings 
demonstrate that soil salinity can 
be appraised and mapped without 
need for collecting any soil 
samples or for carrying out 

laboratory analyses by using sensor-measurements of soil electrical conductivity (by any of the 
three methods tested: four-electrode probe, four-electrode surface array, and electromagnetic 
induction) and simple estimates of soil water content relative to field-capacity and clay percentage 
made in the field by "feel" methods. The simpler, "soil-type" calibration-approach was deemed 
unsuitable for this situation where the soils varied so much in texture and moisture condition 
within individual fields and from one field to another over the area. The variability of soil-type 
within mapping units was concluded to be too large for this latter method to work well in this 
situation. However, the EC -model approach gave good results, irrespective of this variability- 
problem. 




90 180 270 
X, metres 




90 180 270 360 
X, metres 



More recently, soil salinity has been characterized in even more detail than that described 
above using the mobilized four-electrode and EM systems and the newer "stochastic field- 
calibration" technique for converting EC a readings to EC e described earlier. Numerous fields have 
been successfully surveyed with the mobile systems, collecting readings at spacings that provided 
a grid-like pattern of required/desired intensity (generally between 10 and 50 m apart). The 
locations of the measurement and calibration-sample sites were established using the GPS 
technology described above. A small number of the measurement sites were selected for soil 
sampling (ground-truthing) based on the observed EC a field pattern (using the ESAP software 
described in Chapter 3, section Determination of locations of measurement and calibration 
sites). A second trip into each surveyed field was undertaken to collect the relatively small 
number (usually 8-16) of soil samples using rapid, tractor-mounted augering/coring equipment. 
The salinities of these soil samples were then determined using the rapid method of Rhoades et al. 



Soil salinity assessment 69 



(1989b). The salinities at the remaining nonsampled sites were predicted from the corresponding 
EC-sensor readings through use of the multiple linear regression relation (Equations [33] and 
[34]) established for each field using the ESAP software. The salinity contour-maps obtained 
using this method/software were nearly identical to those obtained by conventional soil 
sampling/laboratory analysis and cokriging methodology. An example of this approach is 
illustrated in Figure 58, after Lesch et al. (1992) and Rhoades (1997b). This approach provides a 
very practical and cost effective means to substantially reduce the number of soil samples needed 
to accurately map the detailed spatial salinity patterns that occur at the field scale (see Chapter 
5). Such detailed spatial data can also be used to assess the adequacy/appropriateness of 
irrigation/drainage systems as is discussed later. This methodology is less suitable to map large 
areas in broader detail, such as the variable 15 -square mile area discussed just above, because 
each field would have to be surveyed in full detail. Thus, for the less detailed broad-scale 
mapping purposes the "EC-model" technique is recommended; the "stochastic field-calibration" 
technique is recommended for detailed field-scale mapping purposes. 



Monitoring Soil Salinity 

It is important to be able to detect changes and trends that are occurring in salinity conditions and 
patterns over time in fields and projects, in order to be able to detect emerging problems, to 
evaluate the effectiveness/appropriateness of management practices, especially newly 
implemented ones, and to determine the progress of reclamation efforts. Traditional statistics 
provide tests to compare the means of two populations for differences and can be applied for 
some of these needs. However, the changes in the spatial levels and distributions of salinity within 
the soil profile and within the various fields of the project are also of interest. Formal quantitative 
statistics for monitoring purposes compatible with the instrumental salinity assessment 
technology described herein have been lacking until just very recently (Lesch et al., 1998). The 
methodology developed in the latter research utilizes the stochastic field-calibration technique of 
predicting salinity from EM-38 and/or four-electrode sensor measurements and a field-specific 
calibration based on limited soil-sample data. It is advocated herein, together with certain test- 
statistics for monitoring purposes, i.e., for determining if the salinity pattern of a field has 
changed or if the average salinity level of the entire field has changed over time. This 
theory/methodology was successfully tested and its utility demonstrated using the data of Diaz 
and Herrero (1992) and Lopez-Bruna and Herreo (1996), which is rather unique, in that it is one 
of the few published data sets where both EM-38 and soil salinity data were both acquired at 
multiple times from within the same field. Because the statistics involved in this procedure are 
rigorous, as well as beyond the training expected of the typical reader, they will not be presented 
here in detail. Rather, the interested reader is referred to the publication of Lesch et a/.(1998) for 
this information and to Annex 9 where a very brief description of the test-statistics and relations 
are provided. However, a brief qualitative overview of the approach and tests will now be given 
to illustrate the general features of the monitoring approach and methodology. 

The basic approach used to monitor soil salinity is: (a) first, to estimate a regression model 
(using Equations [33] and [34]) which is capable of predicting soil salinity at every grid site 
within the surveyed field from the collected sensor readings, (b) at some future time, to acquire 
new soil samples at two or more of the previously surveyed sites, and (c) finally, to apply the 
formal test-statistics described in Annex 9, in order to determine if differences exist between the 
reviously predicted and recently observed salinity levels and patterns. Two test-statistics are used 
in this comparison: (a) a test to detect a change in the fields median value of salinity between the 
initial and current times and (b) a test to detect any change in the dynamic spatial variation of the 



70 Example uses of salinity assessment technology 



salinity pattern over time - i.e., to see if the salinity pattern has changed in a non- random, 
dynamic manner across the field. Neither of these tests require that the entire field be resurveyed 
nor that extensive repeated soil sampling be undertaken. For example, 15-20 soil samples are 
usually sufficient for monitoring purposes for establishing the initial spatial regression equation 
for a field; the acquisition of 8-10 new samples are typically sufficient for each subsequent period 
of testing. However, if a new map is desired in order to display the new pattern, presuming it has 
changed, a second full-survey of the field is required, as well as the development of a new 
regression model between the sensor readings and soil salinity appropriate for each successive 
testing period. Although the second set of survey data is not required to compute the test 
statistics, these data must still be acquired in order to create a new salinity map. Since, the 
changes in the pattern of salinity in a field are generally meaningful and of interest, it is 
recommended that the analyst repeat both the sensor readings and the calibration procedure for 
each subsequent time period for which significant changes in salinity condition have been 
detected. Whenever these surveys are to be repeated, the successive survey-sites should be co- 
located with the initial ones in order to permit the test-statistics to be correctly applied. One 
should not try to apply the predictive regression relation established between the sensor readings 
and soil salinity for the first survey time to any subsequent time, nor should one try to apply the 
test-statistics to data sets that were established on non overlapping grids. 

A more general treatise on various statistical methods available to determine and map soil 
degradation over time is that of Hoosbeek et al. (1997). 



Developing Information for Site-Specific Management 

Information concerning the spatial distribution of various soil physical and chemical properties 
within a field are needed, along with correlated plant yield relations, to optimally develop a site- 
specific farming plan, i.e., management that accounts for the variability of soil properties and 
crop differences that exist within the field. Among the soil properties of interest, besides salinity, 
are: infiltration rate, water-holding capacity, drainage rate, micro-relief, soil-depth, fertility, 
organic matter content, pH, and texture. Traditionally many of these soil properties have been 
estimated and mapped from laboratory analyses of soil samples collected on the basis of a 
relatively coarse grid, due to the lack of practical ways to measure them directly in the field. 
Alternatively, yield maps have been developed by spatial-samplings as a means to estimate the 
different crop input needs that vary among the various regions of the field. Correlations between 
these soil properties and plant responses are being sought to identify the causes of observed 
spatial-differences in yield and to develop predictive models for estimating the spatially varying 
farming input needs that exist within individual fields or management units. Management to 
compensate for field variability in salinity has not received much attention in the past, but, the 
author believes, it will in the future. 

A limitation in the use of conventional soil sampling/laboratory analysis methodology for 
characterizing the spatial variability of soil properties is the high labor requirement involved. 
Typically for prescription farming applications, a grid of 100 by 100 metres (330 by 330 ft) is 
used (about one sample per ha. about one sample per 2.5 acres), which often is not intensive 
enough. The proper grid spacing depends upon the variability of the property of interest, which, 
of course, is unknown at the outset. Thus, the proper locations to collect the samples and the 



Soil salinity assessment 



71 



number of samples required cannot easily be determined by the conventional approach. As a 
result, too few samples are frequently taken to properly characterize the variability that often 
exists in fields for prescription farming purposes. No cost effective, scientific approach for 
determining grid size has been developed using such grid-point methods. Thus, directed sampling 
and remote sensing techniques are being sought and advocated, in order to site optimum soil-test 
locations and to minimize sampling needs. But traditional methods of directed sampling and 
remote sensing often do not provide enough, or sufficiently quantitative, information about the 
various soil properties described above for the needs of prescription farming. 

On the other hand, measurements of EC a and of geospatial position can be obtained 
rapidly with geophysical sensors and used to determine optimum soil-test sites as explained 
earlier; additionally, EC a can be used to infer a number of soil properties, besides salinity, that 
are useful to prescription farming purposes and thus to create much more detailed and affordable 
soil-property maps than those obtained by the use of conventional soil/grid-point sampling 
methods (Kachanoski et ah, 1988; Lesch et al., 1992; Doolittle et ah, 1994; Jayne, 1996). The 
theory for using measurements of soil electrical conductivity and spatial calibrations of pertinent 
soil properties for use in prescription farming applications is described in the paper by Rhoades 
et al. (1997d). Some example-applications for salt-affected soils follow. 



FIGURE 59 

Three-dimensional map of the electrical conductivity of bulk soil (EC a ) of a salt-affected field 

in the Coachella Valley of California USA (after Rhoades et al., 1997d) 



LEGEND 



0.1 
0.9 
1.6 
2.3 
3.1 
3.8 
4.5 




SW 



There are numerous situations where the reclamation and management needs of saline soils 
vary within individual fields and prescription farming methods could be used to advantage. An 
obvious situation is evident in Figure 59, which displays the marked variability of EC a that 
existed in a salt-affected field located in the Coachella Valley of California, as measured by the 
mobile, combination four-electrode/EM system previously described. These data were converted 
to units of soil salinity by the stochastic field-calibration calibration method and corresponding 
maps were prepared. The results showed that salinity was exceedingly excessive (see Figure 60) 
for crop growth in most of the field; the median EC e level was 43 dS/m and ranged from 3 dS/m 
to 106 dS/m. Additionally, soil sodicity, as expressed in terms of the sodium adsorption ratio of 
the saturation extract (SAR e ), was very high (see Figure 61), and quite variable, in the field 
(median value of 146; ranging from 9 to 475). 



72 



Example uses of salinity assessment technology 



FIGURE 60 

Estimated salinity (EC e in dS/m) of the 0-60 cm depth- 
interval of a salt-affected field in the Coachella Valley of 
California USA (after Rhoades etal, 1997d) 



331, 




■ 


ECr >fi0 


B 


4*«ECi*6» 


3 


K>*EC<4[] 


n 


EQ,*20 



:;;* 



FIGURE 61 

Estimated sodicity (SAR e basis) of the 0-60 cm depth- 
interval of a salt-affected field in the Coachella Valley of 
California USA (after Rhoades etal, 1997d) 



SAR e is generally well 
correlated with EC e in most fields; 
as is the case here, though less so 
than typical (r = 0.78). 
Reclamation was obviously re- 
quired before this field could be 
successfully farmed and a plan 
was sought accordingly. In this 
regard, it was recognized that the 
reclamation of such saline/sodic 
soils may, or may not, require the 
use of amendments to replace 
exchangeable-sodium and to 
sustain permeability during 
leaching. To determine the spatial 
differences in the needs of gypsum 
for reclamation purposes in this 
field, the combinations of 
SAR e /EC e existing in the field, as 
obtained by the detailed 
measurements of EC S and the 
analysis of soil samples collected 
from the "calibration" sites 
selected using the spatial- 
sampling procedures described 
earlier, were classified into four 
categories. The spatial patterns of 
these four conditions of SAR e / 
EC e are depicted in Figure 62. 
Bulk samples of soil (0-60 cm 
depth) were collected from these 
four regions, packed in 
permeameter-columns, with and 
without the addition of gypsum 
(2, 5 and 10 tons per acre basis; 
4.5, 11.2 and 22.4 Mg per ha.) 
and leached with the local 
irrigation water. The hydraulic 

conductivities of these soil-columns were monitored throughout the more than five pore- volumes 
of leaching they were subjected to, as were the EC and pH levels of the effluents. The results 
obtained showed that only the regions of the field with SAR e /EC e ratios of greater than 5 and EC e 
levels of less than 20 dS/m benefited from the addition of gypsum; without it the permeabilities of 
the soils in these areas decreased by more than a factor of two after 2 pore volumes of leaching, 
as the effluent EC decreased below 2 dS/m and its pH increased to 9.3 or greater. Given this 
information, it was determined that gypsum would be beneficial in only a small area of the field, 
that part of the field shown in Figure 62 having SAR e /EC e ratios greater than 5. In this manner, 
the reclamation prescription, in terms of gypsum requirement, was established for the different 
classes of chemistry and soil type existing in the different regions of this saline/sodic field. 







Soil salinity assessment 



73 



The overall need for 
gypsum determined by this 



FIGURE 62 

Estimated salinity/sodicity ratio of the 0-60 cm depth- 
interval of a salt-affected field in the Coachella Valley of 
California USA (after Rhoades et al., 1997d) 



ME 



sm 



E 



spatially discriminating proce- 
dure (prescription farming 
approach) was less than one-third 
of the amount that would be 
required to uniformly treat the 
field, as is traditionally done. The 
resulting savings in the cost of the 
gypsum and its application was 
more than US$ 25 000 for this 
one relatively small field (40 
acres; -16 ha.). The conditions of 
salinity and sodicity existing in 
salt-affected soils are typically 
spatially variable, not unlike those 
seen here. Thus, one can conclude 
that the reclamation requirements 
of typical salt-affected fields are 

spatially variable and can be defined and prescribed advantageously using the general approach 
undertaken in this example. 



i« 



a-i 





■j 


^^^ 




W-lv.-vV-V ... OPC ■£ 


Bf.jj 




^^ . 




• '. ' ' ■"> ?*mmrJbH^ 




"'^illlP 










J^g^S 






_'jii'l'j( 



■ 


n.ii .'■ 


a 


I •;.[.• . ■:. 


z 


Rito>9 


□ 


i ll.- ■ •; 



1SJ 
X, metres 



21* 



FIGURE 63 

Three-dimensional map of the electrical conductivity of bulk soil (EC a ) of a salt-affected 

Sudangrass field in the Coachella Valley of California USA (after Rhoades et al., 1997d) 



LEGEND 

1.4 — 
1.9 — 
Z.4 

Z.9 — 
3.4 
3.8 
4,3 *— 



EC ,dSAn 




SE 



NE 



Another salt-affected field in the Coachella Valley of California that was "surveyed" with 
the mobilized EC a measuring equipment is depicted in Figure 63. Like the other salt-affected field 
discussed above, this field also displayed substantial spatial variation in EC a (and 
correspondingly in salinity and sodicity). However, in contrast to the previous field, the pattern in 
this second field was markedly cyclic in a north-south orientation, as illustrated in Figure 64; the 
"valleys" in the north-south oriented traverses correlated spatially with the presence of a sub- 
surface tile-drainage system (which will be discussed more later). 



74 



Example uses of salinity assessment technology 



FIGURE 64 

Relationship between the electrical conductivity of bulk 
soil (EC a ) and distance across (and perpendicular to the 
sub-surface drainage system) a salt-affected sudan- 
grass field in the Coachella Valley of California, USA 
(after Rhoades, 1996b) 




10D 



200 3D0 

L«i*ri* piMwioe, «wm*i 



400 



The soil salinities (EC e 
basis) corresponding to the EC a 
values shown varied from 
relatively low values of 2-3 dS/m 
in the regions of the field 
overlying the drain lines to 
relatively high values of 20-25 
dS/m in the mid- line regions. The 
height of the sudan crop planted 
in this field was well correlated 
with the salinity/drainage pattern, 
as is illustrated in Figure 65. The 
spatial variation in relative sudan 
yield (assuming it is proportional 
to height) predicted from the data 
shown in Figures 63 and 65 is 
shown in Figure 66. These results 
lead to the conclusion that the 
drainage system in this field is 
inadequate, either because of 

"clogging" or insufficient capacity for the given situation of irrigation practices and local 
hydrology. In any case, they indicate that different parts of the field vary in their management 
needs. For example, the areas of high salinity have less need for fertilizer, because of the reduced 
crop growth there, and greater need for effective leaching and drainage. The spatial variation in 
input needs and management requirements existing in this field is conducive to the 
implementation of prescription farming methodologies. Management should be altered to 
increase the rate of drainage and the extent of leaching in the "midpoint" regions of the field, 
either through renovating the drains or decreasing their spacing. Meanwhile, the amount of 
fertilizer applied should be reduced in these areas and the seeding rate increased. Other 
management practices to reduce the level of salinity in the seedbed and to improve irrigation 
efficiency, as described elsewhere (Rhoades et al., 1992), should also be adopted in these areas. 

A third field, this one from the Imperial Valley of California, with excessive salinity in 
certain sections of the field, as determined by calibrated spatial measurements of EC a , is 
illustrated in Figure 67. In this case, the salinity increased with distance along the "head- to-tail 
traverses made across the field (only one traverse is shown in Figure 67) and was excessive in the 
"lower-third" region of the gravity, furrow-irrigated field for the full-production of even the salt- 
tolerant sugarbeet crop growing there (this pattern is commonly observed in fields irrigated by 
such irrigation methods). The distributions of salinity observed within the root zone (data not 
shown) indicated that excessive water had been infiltrating the "upper" sections of the field, while 
inadequate amounts had been infiltrating in the "lower" sections (which will be discussed more 
later). Obviously this situation creates variable management needs for the differentially 
irrigated/salinity-affected regions of the field. Management needs to be altered to improve the 
uniformity of irrigation and infiltration in this field. 



The latter two examples illustrate a form of prescription farming that is little mentioned 
and utilized, i.e., management to accommodate or mitigate the non-uniformities that occur in 
water application, crop-consumption, leaching and drainage in gravity-irrigated fields. 



Soil salinity assessment 



75 



FIGURE 65 

Relationship between the EM-38 sensor readings of EC a , Sudangrass plant height, and 
distance across (and perpendicular to the sub-surface drainage system) a salt-affected 
Sudangrass field in the Coachella Valley of California, USA (after Rhoades et al., 1997d) 



E 



(3 



1.75 
1.50 
1.25 
1.00 



"3 0.75 



0.50 
0.25 



-> — i — ' — r 



observed height 
predicted height 




O (BO 



12 3 4 

EMv . dS/m 



5 25 50 75 100 125 150 

Distance, metres 



FIGURE 66 


Three-dimensional map of the predicted Sudangrass yield (in terms of height) predicted from 


the data of Figures 64 and 65 for a salt-affected Sudangrass field in the Coachella Valley of 


California USA (after Rhoades era/., 1997d) 


LEGEND: CROP HEIGHT (m) 


03S ^ _^Mtib 


0.51 /■>_■' B^ ^r^^^^i^jiM 


065 — ^ ^0^ 


OS* — iS 
LOO — ^^^ 


L14 — ^ W 


13,1 — *!^ ^T 


M ^r m 






SE HE 



76 



Example uses of salinity assessment technology 



The use of the spatial 
measurements of EC a and of the 
spatial-methods of their calibration/ 
interpretation, as described and 
advocated herein, offer a potentially 
valuable new tool to assess and better 
manage such variable-field/soil 
situations using prescription farming 
approaches. More details about the 
variability of salinity in the fields used 
in these two examples, and in others, 
are given elsewhere (Rhoades, 1992b, 
1994, 1996b; Rhoades et al., 1997a, 
1997b). More details about the use of 
spatial measurements of soil salinity 
and electrical conductivity for 
prescription farming purposes is 
described in Rhoades et al. (1997d). 



Evaluating Adequacy and Appro- 
priateness of irrigation/drainage 

As mentioned earlier, the distribution 
of salinity within the root zone of a soil 
is a reflection of the direction of the 
past net flux of water flow. A net 
upward flow, such as may occur in the 
presence of a shallow water table or 
otherwise poorly drained situation, is 
reflected by the presence of high 
salinity in the near-surface depth of 
soil and by decreasing levels with 



FIGURE 67 

Relationship between (a) soil electrical 
conductivity (EC a ), as measured by both the 
mobilized, four-electrode system and the 
mobilized, electromagnetic (EM) system, and (b) 
measured and predicted soil salinity (EC e basis) 
and distance along a transect across a furrow- 
irrigated, sugar beet field located in the Imperial 
Valley of California, USA (after Rhoades, 1997) 



V) 

•o 



3 



o 
V) 



14 



e 

W 12 



a 
to 



10 



lour-eleclrode 1 j JkM 


(b) 

1*, fly p \ 1 n 


' ' '♦To 


1 MTrd fd <rji *<jl 


four-electrode predictions 

° EM-38 predictions 

* measured EC* level 


.,,.,,,,,,,,, 


.1 > i > 1 l.j uj^XuuouX 



120 180 240 

Lateral Distance, metres 



300 



360 



depth to a minimum level determined by the salinity of the shallow groundwater. On the other 
hand, a net and excessive downward flux of water through the soil is reflected by low levels 
(controlled by the salinity of the irrigation water) of salinity in the near-surface depth of soil with 
relatively little increase in the deeper depths. Evidence of the credibility of this generalization was 
presented earlier for the saline seep situation. Other examples for the case of irrigated soils will 
now be given to further demonstrate the utility of the salinity assessment technology to evaluate 
the adequacy/appropriateness of irrigation and drainage systems and practices. 



The data obtained with the mobile, four-electrode and EM sensing systems presented in 
Figure 67 can be used to demonstrate this utility. Average rootzone soil salinities, expressed in 
terms of EC e , as predicted from measured EC a data (Figure 67a) obtained in a furrow irrigated, 
sugar beet field (Glenbar silty clay loam soil) in the Imperial Valley of California and as 
measured in some "calibration" samples collected along the transect are shown in Figure 67b. 
The salinity values were predicted from the sensor readings and limited calibration information, 
using the "stochastic field-calibration" technique. As is shown here, the accuracy of these 
predictions is very good. If irrigation application and infiltration were uniform across the field 



Soil salinity assessment 



77 



involved with Figure 67, the value of EC a (and of EC e ) should be the same at each distance, 
provided crop stand and soil type were also uniform. However in this case, the EC a (and EC e ) 
values increased from the "head" to the "tail end" of the field; the coefficient of variability (CV) 
was 14.2% and the linear correlation coefficient (r) between EC a and distance down the transect 
was 0.67. Thus, one may conclude from these data that the field is not uniform with respect to 
one or more of the three possibilities. In this case, the crop was planted uniformly and the soil 
type was essentially the same along the transect. Hence, these observations/data imply that 
irrigation application, or infiltration, was not uniform across this field, presumably due to 
reduced opportunity-time and infiltration of irrigation water with distance from the point of water 
delivery to the furrows. Another factor likely influencing the salinity distribution within this field 
is the lateral transport of salt that occurred in it as a consequence of the "cracking" type of soil 
present in the field. This latter aspect is discussed elsewhere (Rhoades et al., 1997b). This 
example illustrates how information about the spatial variation of average root zone soil salinity 
can be used, assuming it is a tracer of the interactions of water infiltration, evapotranspiration, 
leaching and drainage, to evaluate irrigation uniformity in fields which are relatively uniform in 
soil type and cropping intensity. 



FIGURE 68 

Correspondence between soil salinity predictions based 
on soil electrical conductivity measurements obtained 
with the mobile salinity assessment system along a 
transect across a surface irrigated, tile-drained alfalfa 
field (Imperial clay soil) located in the Imperial Valley of 
California (after Rhoades, 1996b) 



e 

o 

i 

O 

HI 



VI 

o 
S 1 



'5 



13 



11 



7 - 



5 



I I ■ ■ ■ I I I III — TT — r-i — r-i — T^ — ' — ■ | ■ I ■ I , ■ I I | 

CornparEson of EM-38 and Four-Sflctrode date 
Imperial Clay Soil 



Another example of 
spatial-data obtained in an 
irrigated/drained field with the 
mobile, four-electrode sensing 
system to a depth of 1.2 metres 
is presented in Figure 68. This 
figure shows EC e values 
calculated using the stochastic 
field-calibration method from 
EC a readings collected every 
second (about every 1 m apart) 
as the tractor moved across a 
furrow irrigated, tile-drained 
alfalfa field (Imperial clay soil) 
in the Imperial Valley of 
California. The "minimum" in 
the EC e readings occurring at 
about 380 metres from the 
irrigation-intake end of the field 
corresponds to the position of a 
suite of subsurface drains. 
Otherwise, the EC e values 

increased toward the "tail end" of the field, presumably due to reduced application and infiltration 
of irrigation water with distance from the point of water delivery to the furrows and to lateral 
transport of salt across the field as a consequence of the soil cracking and lateral- transport 
phenomena previously mentioned (Rhoades et al., 1997b). Examples of fields with much greater 
increases in "tail end" salinity have been observed in other fields (Rhoades, 1992b). Tile drains 
are also located in the tail end of the field involved in Figure 68, but were ineffective in lowering 
the salinity except in the four narrow regions, indicated by the sharp downward spikes apparent 
in the "curve", where the trenching had occurred many years (decades) earlier. These data 
suggest that much of the variability in average root zone salinity observed in typical 
irrigated/drained fields is caused by the interactive, effects of the drainage and irrigation systems 
and trenching operations. They also demonstrate how the adequacy or inadequacy of a drainage 




F«jrEI«lroit« dslt 

3 BJ-JSdila 



1 ■■■-.. ■ ■ L , ■ ■ - I ■■■■'■■ 



1M 



MO 3M 400 SCO 
I arpral nklannA rnntrns 



EGO 700 



78 



Example uses of salinity assessment technology 



FIGURE 69 

Relationship between soil electrical conductivity (EC a ) 
and distance along a transect crossing two sets of 
subsurface tile-drains in a salt-affected field located in 
the Coachella Valley of California (after Rhoades, 1994) 

Smoothed Mobile Four-Electrode Data: 
Kohl Farm, Coachella Valley 



6 

Hi 



O 

O 

a 

u 
a 
w 

□ 
to 




27.5 



362.7 



system can be inferred from the 
detailed spatial salinity 

information provided by the 
mobilized salinity assessment 
systems. 

Another example of the 
marked effect that a subsurface 
drainage system can have on 
average root zone salinity is given 
in Figure 69 in terms of EC a . 
Corresponding values of EC e (not 
shown) cycled between low values 
of about 2.5 dS/m to high values 
of about 25 dS/m. The CV and r 
values for this EC a - distance 
traverse were 36.8 % and -0.20 
respectively. This example in- 
volves a field of silty loam soil in 
the Coachella Valley of California 
which has two sets of buried "tile- 
lines" oriented perpendicular to the 
direction of the EC, traverse; one 
set being about 2.7 m deep and 
spaced about 90 m apart and 

another set being about 1.7 m deep and located at one-third and two-third distances between the 
deeper lines. The two sets of tile-lines are represented by the solid and dashed lines, respectively, 
in the figure. In this field, soil salinity levels "mimicked" the drainage system, with high values of 
EC a (and EC e ) occurring in the soil located between tile-spacings and low values in the soil 
overlying them. These data suggest that most of the variability in average root zone salinity 
across this field was caused by the effects of the drainage system. They also imply that the 
drainage system there was inadequate given the circumstances of irrigation, soil type, 
geohydrology, etc. The distributions of salinity within the root zone depth of such fields will be 
discussed later; they give further credence to the preceding conclusion. 

The time involved in collecting these data was only about seven minutes. Again, they 
demonstrate the utility of the assessment methods for evaluating the adequacy of the drainage 
conditions of the field. An even more dramatic drainage system effect on soil salinity is evident in 
the previously discussed Figure 64; again, such data show the utility of the rapid salinity 
assessment equipment and methodology. 



til 3 195,1 278.9 

Lateral Distance, mot res 



deep lisa lines 
shale* tile lines 



The spatial pattern (average root zone basis) of the field depicted in Figure 69 is shown in 
Figure 70. The average profile EC e value of 10-12 dS/m measured within the 0-1.2 m depth in 
much of this field is excessive for successful crop production. This observation itself is evidence 
of the inadequacy of the past irrigation and drainage management in the field. Assuming uniform 
irrigation and a leaching fraction of 0.05, the expected value of average root zone salinity (as 
calculated using WATSUIT, Rhoades et ai, 1992) would be about 2.1 dS/m under steady-state 
conditions of irrigation with the Colorado River water. Since the average soil-profile salinity in 
this field of silty-loam soil (non-cracking soil) exceeds 2.1 dS/m, one must conclude that the 
overall leaching fraction is negative either because of deficit-irrigation or because salt is being 



Soil salinity assessment 



79 



FIGURE 70 

Predicted average root zone soil salinity in a tile- 
drained field located in the Coachella Valley of 
California, USA (after Rhoades, 1997a) 



so ;■ 




Average Salinity 
wiinm n.D-1 .a m ueptii 



EO>>12d5fm 
e .. r;;,. .: !■: ::s.' i 

ECiv6c5.ni 



FIGURE 71 

Relationship between salinity distribution and mean 
level of salinity in the soil profile of a tile-drained field 
located in the Coachella Valley of California, USA 

(after Rhoades etal., 1997a) 



accumulated in the root zone from 

the upward flow of saline water 

from the shallow groundwater. 

Since information supplied by the 

irrigator showed that the applied 

water exceeded ET, the latter cause 

is deduced. The salinity 

distributions found within the 

profiles over much of this field are 

presented in Figure 71; they are 

concluded to be affected by the 

drainage system. As discussed 

above, lower salinities occurred in 

this field in the soil overlying the 

tile-lines and higher salinities 

occurred in the soil located in 

between the tile lines. Additionally 

in this field, as shown in Figure 7 1 , 

the distribution of salinity in the 

soil profile varied with the mean 

level of salinity. These distributions 

imply that salinity is high in areas 

where the net flux of water has 

been upward in the field (in the 

region of the field located in 

between the drain lines) and is low 

in the areas where the flux has been 

downward, i.e., where leaching has 

occurred in the soil overlying the 

tile lines. Figure 72 portrays the 

salinity distribution in the upper 

part of the root zone (0-0.5 m) of 

the Coachella Valley field. These 

data indicate that the salinity levels 

and patterns within the seed bed of 

this field are also related to the 

mean profile salinity levels, which in turn are related to the drainage pattern. Taken together, all 

these data (Figures 69 to 72) indicate that the drainage system in this Coachella Valley field is 

inadequate given the manner of irrigation, or geohydrologic situation, or both, existing there. As 

shown in Figure 72, the salinity distributions in this Coachella Valley silty-loam soil are clearly 

two-dimensional in contrast to the one-dimensional profiles observed for the clay textured 

Imperial Valley soil (see Figure 73, after Rhoades et al., 1977b). This latter figure shows that 

salinity in the center of the seed-bed of the fine-textured soil is not as high as might be expected. 

A likely reason for this is the presence of an extensive network of cracks within the bed which 

allowed water movement through it, especially in the later stages of the irrigation season 

(Rhoades et al., 1997b). 



o 
5. 





— I — i — • — i — ■ — i — i — i — ■ — i — ■ — ■ — I — ■ — > — I — 

Typical Salinity Profiles 


o.ts 


" \ \ 


\ y 


0.4G 


■ V w )) 




D.75 


- 




Kohl Farm: 


1.95 


\r 


Coactietla Valley 



14 



20 



28 



32 



Soil Salinity. EC dS/m 



This "inter-flow" likely leached out salts which otherwise would have accumulated by 
capillarity and upward flow and evaporation of water in the bed, if it was completely isolated 
from the furrows. The patterns of salinity within the soil profiles of the Imperial Valley soil were 



80 



Example uses of salinity assessment technology 



FIGURE 72 

Salinity distribution in the soil profiles of a tile-drained field located in the Coachella Valley 

of California USA, as influenced by mean (0-0.5 m) salinity level (after Rhoades et al., 1997a) 



EC* < O.Su 

0.5u < ECe < u 

U < EC* < 2U 

2u < EC* < 4u 

EC* > 4u 

note: u - mean ECe 




■ » ojs a.eo tM Mt 

Lateral offset from centre of bed (m) 




OH OiS o.no OJI 

Lalcrai oifstt from centre o( bed |m) 



Sii AN 3?5 i: 50 

. ■■isi- ::!'.-; e1 from centre of bed (m) 



FIGURE 73 

Average salinity distribution in the soil profiles along a 
transect across a furrow-irrigated, tile-drained alfalfa 
field (Imperial clay soil) located in the Imperial Valley of 
California, USA (after Rhoades et al., 1997b) 



o 
Q 

O 

■li 

Ed 

o 



BSB3E3 



C 



very similar at various points 
along the transect; how-ever, in 
relation to the average profile 
shape, salinity increased in the 
upper part of the profile and 
decreased in the lower part of 
the profile with distance towards 
the down gradient end of the 
furrow-irrigated field (see 
Figure 7 in Rhoades et al., 
1997b). 

Salinity "distribution" 
data obtained with the "com- 
bination sensor system" in two 
other fields (both near each 
other in the San Joaquin Valley 
of California) are given in 
Tables 4 and 5 to further 
illustrate how information about 
the levels and distributions of salinity within the rootzone obtained with the mobilized EC-sensor 
systems can be used advantageously to evaluate the adequacies of salinity control and irrigation 
and drainage management. The percentages of the Borba-farm field having levels of salinities 
with certain ranges are given in Table 4. By reference to salt-tolerant tables, one can use these 



o.«o 



0.20 



0.4O 



0.50 



1.00 



Imperial Clay Soil 


i lllilllblll iii 1 iii 

! 1 . 1 - ! l-|-||-|l||-|| r | | hi 1 i 1 1 1 14 1 II 





EC<B 

8 -a EC* *l 1 

1 o * ECe < i : 

] EC«>12 

EC. expressed 
iiJ3.ni 



D.SD 025 O.flO D.ZS O.ED 

Lndcral off act from centre ot bnl 4mk 



Soil salinity assessment 



81 



data to estimate how much yield loss caused by such salinity conditions would result for any 
given crop. For example, assuming the crop is alfalfa (which has a threshold EC e value of 2.0 
dS/m and a rate of yield loss of 13% for each unit of EC e in excess of 2.0; Maas, 1990) and its 
effective depth of rooting is 1.2 metres, one would estimate the relative yield loss due to salinity 
to be as follows by percentages of the Borba field: 0% loss in 3% of the field, 14.6% loss in 49% 
of the field, 44% loss in 29% of the field, and 100% loss in 18% of the field. Thus, on a whole 
field basis, the expected salinity induced loss in relative alfalfa yield would be 38%. The 
economic significance of this yield loss in turn can be calculated given other cost information and 
used to evaluate the economic impact of salinity on the profit-line of the operation of this field 
and also to evaluate the affordability of improving the management to eliminate the salinity- 
induced yield losses. 



TABLE 4 

Percentages of field with soil salinities (EC e ) within certain ranges 



Soil salinity 
dS/m 


Soil depth, metres 


0-0.3 


0.3-0.6 


0.6-0.9 


0.9-1.2 


0-1.2 


0-2 


14 


44 


17 


15 


3 


2-4 


41 


32 


34 


31 


49 


4-8 


36 


17 


22 


25 


29 


8-16 


9 


6 


16 


17 


16 


<16 





1 


10 


11 


2 



TABLE 5 

Percentages of fields by different soil salinity (EC e basis) profile types 



Profile characterization 


% area 


Salinity Profile Ratio 


Profile Type 


Furrow-field 


Sprinkler-field 


>0.75 


very negative leaching 


5 





0.50-0.75 


negative leaching 


23 


3 


0.35-0.50 


excessive leaching 


17 


13 


0.20-0.35 


normal leaching 


35 


71 


<0.20 


low leaching 


20 


13 



As explained earlier, the information of salinity by depth and location in the soil profiles of 
irrigated fields, as is provided by the mobilized EC-38 sensor system, can be used to assess the 
adequacy of the past leaching and drainage practices. For example, where salinity is high in the 
near- surface soil of non-deficit irrigated fields and decreases with depth in the profile, the net flux 
of water (and salt) can be inferred as having been upward. This is reflective of inadequate 
drainage. Where salinity increases with depth in the profile, the net flux of water and salt can be 
inferred as having been downward. When salinity is low and relatively uniform with depth, 
leaching can be inferred as having been excessive, probably contributing to water-logging 
elsewhere. As shown previously (Table 29 in Rhoades et al., 1992), an approximate relationship 
can be established between steady-state leaching fraction (L) and the ratio: EC e in the top-half of 
the root zone/the sum of EC e throughout the profile. This relationship (see Figure 74) between L 
and the latter ratio (salinity profile ratio, P) is: L = 0.01843 (e 80P ). Thus, one can infer the 
approximate degree of leaching (under steady-state conditions) from the salinity profile ratio, 
which, in turn, can be determined from the data acquired with the mobilized EM-38 sensor 
system. As an example, the percentages of a furrow-irrigated cotton field in the San Joaquin 
Valley of California are given in Table 5 by classes of salinity profile ratios. Inverted salinity 
profiles (P > 0.50) occurred in 28% of this field. Such profiles are indicative of the net upward 
flux of water for the reasons previously given. The author speculates, knowing that the irrigator 
applied water in excess of ET in this field, that excessive deep percolation occurred in the pre- 
season and early-season irrigations, causing a "mounded, perched" water table which was the 
source of the water and salt that subsequently "subbed" back up into the root zone. Profiles with 



82 



Example uses of salinity assessment technology 



salinity distributions indicative of 
excessive net-leaching (L values of 
greater than 0.3; salinity profile 
ratios of 0.35-0.50) occurred in 
17% of the field, and profiles with 
salinity distributions indicative of 
normal leaching (L values of less 
than 0.3; salinity profile ratios 
<0.35; salinity increasing with 
depth) occurred in only 55% of the 
field. Such data indicate that the 
leaching/drainage management this 
field has received was inadequate 
over much of the field. The 
analogous percentages obtained in a 
nearby San Joaquin Valley field 
(this one sprinkler irrigated) are 
also given in Table 5. While both 
fields were of the same soil type 
(SiCl) and water table depth (-1.5 
m), quite different results were 
obtained. Hardly any of the 
sprinkler-irrigated field had inverted 
(net upward-flux; P>0.5) profiles; 
the desired normal leaching profiles 
were evident over 84% of the field. 

These examples show that 
improved irrigation, drainage and 
salinity management that can result 
from the use of the more efficient 
and uniform method of sprinkler 
irrigation compared to furrow- 
irrigation. These data further 
illustrate the utility of the mobilized 
assessment system and of detailed 
spatial information of soil salinity 
and its distribution through the root 
zone to evaluate the adequacy and 
effectiveness of irrigation and 
drainage systems and practices. 
Maps of the areas with excessive 
leaching or of inadequate drainage 
can easily be prepared from these 
data to display the areal extent and 
locations of these conditions. An 
example of such an application is 
given in Figure 75, which shows the 
pattern of average soil salinity in 



FIGURE 74 




Relationship between the salinity profile ratio 


and 


leaching fraction (after Rhoades etal., 1997a) 




1.1 

i.o 

2 0£ 


L = 0.01 843(es °p) 




1 0.7 






I 0.6 

| 0,5 

3 °* 






| D.3 






8 M 






0,1 






D.I OJ 0.3 0,4 


04 


TH«»9tl£l!tSBllnHy Prolll* FUtlO. P 





FIGURE 75 

(A) Predicted soil salinity (EC e ) and (B) salinity profile 
ratio for a tile-drained field located in the Coachella 
Valley of California, USA (after Rhoades, 1997) 




^verao? Saintly Hap 
0.0-0.S in depth 

■ i2<BCt-t IBdSVm 

a *. re? : 1 2 ds.vi 

G EC<id3/m 



wesl-east 




Salinity Ratio Map 
in :: :: 'ii.lfi :i i .2) 

■ FJH0*»7 

B ftfl -. Rdlki * D.7 

[?3 OS -e Ftalia < 0£ 

1.1 Hallo cO.S 



wfl5l-aast 



Soil salinity assessment 



83 



the to 60 cm. depth (Figure 75 a) of an irrigated field (Kohl-5) located in the Coachella Valley 
of California and the corresponding pattern of the salinity profile-shape ratio (Figure 75b) 
predicted from the calibrated sensor readings. Profile-ratio values exceeding 0.5 imply inadequate 
drainage (a net upward flow of water) for steady-state conditions (the interpretation of the 
profile-ratio is discussed more later). These data imply that salinity in the major rootzone depth 
(0-60 cm.) is high in the areas of the field which are the least well drained, in fact in regions 
where the net flux of water through the root zone over the past years of cropping has been 
upward. 



FIGURE 76 

Cyclic pattern of soil electrical conductivity (EC a ) 
across a succession of furrows, some of which were 
trafficked by a tractor (V) and some which were not 

(after Rhoades etal., 1997a) 



EMU DalJ 



£ 



Besides irrigation and 
drainage, tillage and tractor 
traffic-patterns have also been 
deduced from the intensive, 
spatially referenced data sets 
obtained using the mobilized EC- 
sensor systems to significantly 
affect soil salinity levels and 
distributions in some fields. 
Tractors typically move through 
the fields in a systematic way, as 
dictated by the invoked practices 
of seed-bed/furrow preparation, 
cultivation and tillage. As a result, 
tractor weight is repeatedly 
exerted in some furrows, but not 
in others, leading to cyclic 
patterns of compaction among 
some sequential sets of 
neighboring furrows. Similarly, 
tillage and cultivation operations 
are often implemented using 
equipment with guide/depth 
wheels which similarly lead to 
other analogous definable 
"traffic" patterns. As a result, 

some furrows can become more compacted than others leading to reduced water-intake rates and 
to relatively increased lateral water flow rates and, hence, to higher salinity levels in both the 
associated furrows and beds. Systematic, cyclic differences have been observed in the salinity 
patterns of some irrigated fields surveyed with the mobilized EC-sensor equipment and to 
"mimic" the traffic patterns undertaken with the tillage equipment. An example is shown in 
Figure 76, in which the EC a readings obtained in a succession of neighbouring furrows are 
presented (Figure 76a). The furrows in which the tractor tires traveled are indicated by a small 
inverted triangle. The EC a values associated with the "spline-fit" (the plot of the "running 
average" of neighbouring values) of the readings are indicated by the dotted line. The differences 
between the individual EC a values for each furrow and its spline-fitted value are presented in 
Figure 76b. These data show that EC a (and, by implication, salinity) is substantially higher in 
each furrow in which the tractor tires traveled compared to its neighboring furrows. They also 
show that ECj (thus salinity) is substantially lower in each furrow that is "sandwiched" between 
"traveled" furrows. The other furrows have EC a values that are only slightly higher, or lower, 
than its neighbors, as would be expected if there was no cyclic pattern or significant difference 




m «□ ao en mo 

Dl»1«ne e acrot* furrow*, melius 



un 



84 Example uses of salinity assessment technology 



between them (that is, if all the furrows were essentially the same in their degree of compaction). 
The observed salinity pattern across this succession of furrows was clearly cyclic in nature and 
related to the tractor traffic pattern that had been followed in the field. In some fields which 
displayed this same phenomenon, the EC e values in adjacent beds of furrow-irrigated fields have 
differed from their neighbors by as much as 4 dS/m, or more. Analogous cyclic patterns of soil 
salinity have been observed in other "surveyed" fields that were caused by deep chiseling actions 
of subsurface tillage operations. In this case, the data obtained led to the inference that water had 
infiltrated and flowed preferentially in the tillage "slits", then flowed horizontally out into the 
adjacent soil causing salinity to be lower in the vicinity of the "slit" compared to the inter-slit soil 
areas (data not shown). In one "surveyed" field which had been "ripped" to 0.5 m with chisels, 
markedly abrupt cyclic patterns of EC a were observed that mimicked the tillage pattern. An 
excavation and detailed examination of the soil profile was made at the cyclic locations where the 
abrupt changes in EC, were measured. This examination revealed (once the topsoil was removed) 
the presence of deep narrow trenches, or cracks, approximately 2.5 cm wide in the soil underlying 
the topsoil mulch. An interesting feature of these "cracks" was that they were full of dry 
aggregates of surface soil that had fallen down into them. Hence, such "cracks" not only provide 
preferential paths for water flow, but as well provide a means for soil particles and associated 
organic matter to "fall" to deeper depths in the soil profile and thus a means by which certain 
pesticides and other relatively immobile chemicals may be transported in soils that is not 
accounted for in classical solute transport theory. This observation would not have been made 
without the use of the detailed spatial measurement system. 

The examples given above indicate how the salinity patterns within fields and root zones can 
be used as tracers of the net interactions of irrigation/infiltration, evapotranspiration and drainage 
to deduce much useful information about the adequacy, uniformity and appropriateness of the 
irrigation/drainage/salinity management. The required spatial data can be practically acquired 
using the mobilized systems of salinity assessment. Ways to obtain more quantitative 
interpretations of the amounts of leaching and its associated salt-loads are discussed in the next 
section. 



Assessing Leaching and Salt-Loading 

While salt-affected soils and waters occur extensively under natural conditions, the salt problems 
of greatest importance to agriculture arise when previously productive soil and water resources 
become salinized as a result of agricultural activities (so-called secondary salinization). The 
extent of salt-affected soil and water resources has been influenced considerably by the 
redistribution of water (hence salt) through irrigation and drainage. 

Essentially the same processes are the root causes of both soil and water salinization 
(Rhoades, 1997a). Salinization comes about primarily as a result of the processes described in the 
following scenario. Water containing salt is applied in excess of that which the crop can use and 
the soil can retain, in at least in some of the irrigations and/or in some parts of the field. The 
excess water passes beyond the root zone as drainage flow containing most of the applied salts in 
a reduced volume and proportionately increased concentration. This water, together with that 
percolating downward from canal seepage, dissolves additional salts (over and above those 
present in the applied water) from the soil and underlying substrata. Such concentrated and 
additionally mobilized salts, when transported to lower-lying landscapes and receiving waters, 
result in excessive salt accumulations in these areas, i.e., in soil degradation and/or in water 
pollution. From the preceding it is clear that the major source-areas of salt-loading in irrigated 



Soil salinity assessment 85 



lands are the regions where the application of irrigation water is high, the leaching fraction is high 
and the substrata contains readily dissolvable salts within them. In order to determine the salt- 
loading coming from the root zone, one needs to be able to measure, or calculate, the volume and 
concentration of deep percolation. As mentioned earlier, the protection of our soil and water 
resources against excessive salinization, while sustaining agricultural production through 
irrigation, requires the ability to determine the areas in irrigation projects and in individual fields 
where excessive deep percolation is occurring, i.e., where the water- and salt- loading 
contributions to the underlying groundwater and surface water are coming from (a suitable means 
of determining areal sources of pollution). Though less critical, it is also useful to know the 
amounts of leaching and associated salt-loading. 

As explained earlier, the level and distribution of salinity in an irrigated soil is a reflection of 
the net interactions of irrigation, evapo transpiration, leaching and drainage and, thus, may be 
used as a tracer in this regard. The relationships that have been developed between soil salinity, 
leaching and salt-loading, with reference to the soil profile, will be discussed in this Section. 
These relationships have been based primarily on simple salt- and water-balance concepts, though 
some refinements can and have been attempted to adjust them as needed to account for certain 
deterministic processes. The following equation describes the major inputs and outputs of salts 
that are involved in the salt-balance of the rootzones of cropped soils: 

salt input = salt output + A soil salinity , [35a] 

V iw C iw +V gw C gw +S m +S f =V dw C dw +V tw C lw +S p +S c ±AS sw , [35b] 

where V iw , V gw , V dw and V,„ are the volumes of irrigation water applied, groundwater influx 
either by capillarity or direct invasion of the water table, deep percolation of drainage water, and 
surface runoff (tailwater), respectively; C iw , C gw , C dw and C tw are the soluble salt concentrations 
in the preceding four types of water, respectively; S m and S f are the amounts of salts brought into 
soil solution by mineral weathering and from the dissolution of fertilizers and amendments, 
respectively; S p and S c are the amounts of salts precipitated out of solution in the soil and 
removed from the soil solution by crop uptake, respectively and AS S „ is the change in soil solution 
salinity (Kaddah and Rhoades, 1976). This relation can be simplified (approximated) for certain 
situations by making some reasonable assumptions. For example, except for heavy- textured, 
cracking soils, it may be assumed that C iw and C,„ are essentially the same. It may also be 
assumed, with the shallow water table situation in mind, that C dlv and C,„ are about the same. 
Additionally, because S m , S f , S p and S c are usually small in relation to the other quantities and the 
contributions of the S m and S f inputs tend to offset the losses associated with S p and S c , these 
terms have been traditionally deleted from the calculations (U S Salinity Laboratory Staff, 1954). 
However, S p may be significant and some adjustment for this process may be necessary, if salts 
other than chloride are considered (Rhoades, et al., 1974). Substitution of these "equalities" and 
assumptions into Equation [35b] yields the following relation, where rainfall is insignificant: 

K - y « ) c» + K - v *» ) c *» = A s ™ ■ P6] 

This relation, along with measurements of the volume of infiltrated water (V llv - V, w ), of the 
concentration of the irrigation water (C uv ), of the concentration of soil salinity at the bottom of the 
root zone (Cd w ) and of the change in soil salinity within the root zone over the measurement 



86 Example uses of salinity assessment technology 



period, can be used to estimate the net amount of leaching (Vi = Vd„ - V gw ) and the load of the salt 
(Vi Cd W ) draining from the root zone. 

Substitution of electrical conductivity for concentration and of the volume of infiltrated water, 
Vj„f , for (V iw -V,„) into Equation (36) yields the following equation relating leaching fraction (L, = 
V dw /V in f) for steady-state (AS SW = 0) and good drainage conditions (V gw = 0), as was published in 
Handbook 60 (U S Salinity Laboratory, 1954): 

L f =V dw /V M =EC iw /EC dw . [37] 

For such conditions, the leaching fraction can be obtained from the EC-ratio. Thus, the amount of 
leaching and salt- load can then be calculated, if ECd„ can be measured and if the volume of 
infiltrated water is known. Similarly, the amount of water consumed by evapo transpiration (V cu ) 
can also be determined, since V inf = V cll /(1-L ) under such steady-state conditions. Remember 
that all of the assumptions that are contained in Equation [37] are also inherent in the so-called 
leaching requirement concept, as traditionally applied. Later, the author will give an example of 
the use of this relation and show how a correction can be made for the disparity between the 
concentration ratio and the EC ratio, as well as for the error caused by salt 
precipitation/dissolution processes. 

Rose et al. (1979) derived an analogous pair of relations to Equations [36] and [37], 
respectively, for non-steady-state and for steady-state conditions, meeting the more limiting 
assumptions described below. The intended use of these relations was for predicting from 
relatively short term measurements whether leaching would be adequate in the long term to keep 
salinity within acceptable limits for cropping. The claimed value of this relation was that it would 
permit L (defined below) to be determined from relatively easily obtained information. Retaining 
most of their symbols, the expression for non steady-state conditions given in differential form is: 



z© 



\ 



dt 



= Ic-Ls 7 , [38] 



/ 



where z is soil depth, @ m is mean volumetric soil water content averaged over depth z, s m is the 
mean concentration of a conservative solute (such as chloride) over depth z, t is time (yr.) 
measured from an initial time (to) when s m is first measured, / is the rate of irrigation averaged 
over the time period of measurement, c is the concentration of the tracer solute in the irrigation 
water, L is the leaching flux density at depth z averaged over the time period of measurement and 
s z is the mean concentration of the tracer solute at depth z over the time period of measurement 
at the reference water content m . 

The following assumptions are inherent in Equation [38]: (i) water and solute flow are only 
vertical, (ii) the amount of leaching is low and the soils are only slowly permeable, (iii) the tracer- 
solute is non-adsorbed and non-transformed in the rootzone, (iv) the rates of water application 
and drainage at an arbitrary specified depth z are constant and equal to their mean annual values, 
(v) leaching occurs when the soil is essentially saturated, (vi) there is negligible uptake of the 
solute by the crop and rainfall is negligible, (vii) there is no surface runoff of water, (viii) the 
irrigations and croppings are uniform within the field, and (ix) the shape of the solute-profiles 
changes very little over the time period. 



Soil salinity assessment 87 



Rose et al. (1979) also developed the following expressions for predicting the mean solute 
concentration in the soil profile at later times, including the steady-state condition. For non- 
steady-state conditions, the Equation is: 

S m -S m ^=(lc/Ll-S m ^){l- e x V [-(Ll/z® m )t]}, [39] 

and for steady-state conditions it is: 

s" m =IclLX, [40] 

where s m is the mean concentration of the tracer solute in the soil profile at steady-state and A, = 
s z / s m , which is estimated at any time from the shape of the tracer concentration-profile. The 
history of s m can then be computed with the later two equations for all values of t, assuming the 
irrigation/crop system remains as it was for the period when the data that yielded the value of L 
was collected. The value of the non-steady-state Equation [39] is that it permits the average 
annual value of L to be determined from two measured value of s m and knowledge of / & c over 
time. Equation [40] is equivalent to Equation [37] when X = 1. Equation [38] is analogous to 
Equation [36]. 

Slavich and Yang (1990) modified the equations of Rose et al to account for anion exclusion 
and pore-bypass during leaching. While the refinements contained in their approach to 
compensate for some of the processes which influence leaching-efficiency are physically sound 
and potentially beneficial, it will be difficult in actual field practice to obtain the needed inputs for 
the various parameters, especially for those describing by-pass flow, required by this approach. 
Furthermore, one would not expect leaching efficiency to be single valued but rather to vary with 
irrigation systems, with management, with different soil types, with depth in the soil, and from 
place-to-place in the field. Thus, the method seems needlessly complicated, given the uncertainties 
in the other inputs and is probably impractical. Reasons for these comments will be more 
apparent from the information that is given later in this Section. Therefore, their modified 
relations will not be given herein. 

The above relations (Equations [35] to [40]) provide a means for the estimation of the amount 
of through-drainage (extent of leaching), of the long-term "equilibrium" salinity level resulting 
from irrigation, and of the salt-loading leaving the root zone (or past the maximum depth of 
sampling) in different fields of an irrigation project and within the different areas of individual 
fields. Additionally, they provide a means to locate the primary areal sources of salt-loading from 
irrigation because, as discussed earlier, the amount of salt-loading is proportional to leaching 
volume, though it is also affected by V inf , C iw and the properties of the substrata through which 
the deep percolation flows enroute to its receiving water or soil. Various studies have been 
undertaken to evaluate the above equations; their findings have been reviewed and critiqued by 
Rhoades (1997b) and will not be reviewed here. Most of these evaluations were made using 
analyses (primarily chloride analyses) of soil samples. Such analyses are too time-consuming to 
be very practical for large area assessments of leaching and salt-loading. For this reason some 
attempts have been made to determine if analogous assessments can be made from the more 
practical measurements of in-situ, bulk soil electrical conductivity made using the geophysical 
sensors described earlier, while still utilizing the same salt-balance relations and approach 
(Rhoades, 1980; Cook^a/,,1989; Slavich and Yang, 1990). 



88 



Example uses of salinity assessment technology 



FIGURE 77 

Relationship between electrical conductivity of soil 

water and leaching fraction for the Colorado River 

water used in the Wellton-Mohawk Irrigation and 

Drainage District of Arizona, USA (after Rhoades, 

1980) 



25' 



ECL- Q,8»f0iW5fl A (h O.QOS{ l/L f ) 



r * 0,999 



ix! 






Rhoades (1980) was the first 
to combine the use of EC a measure- 
ments and salt-balance relations to 
estimate leaching amounts in 
irrigated fields. His method is 
practical and will be described to 
show how the assessment techno- 
logy and salt-balance relations 
described in this book can be 
combined into a practical pro- 
cedure for determining the extent of 
leaching and the areal sources of 
salt-loading. He used measure- 
ments of EC a (made at the bottom 
of the root zone with a four- 
electrode EC -probe), along with 
soil-specific calibrations relating 
EC a and EC„ , the assumption that 
the EC of the drainwater is the 
same as the EC of the soil water at 
or below the bottom of the root 
zone, and a modified version of the 
steady-state model (Equation [37]) 
to infer leaching fraction in some 
irrigated alfalfa fields in the 
Wellton-Mohawk irrigation district 
of Arizona. Equation [37] was 
modified to incorporate the effects 
of S m and S p (discussed earlier in 

terms of Equation. [35b]) in the relation between E and the EC iw / ECd„ ratio, as well as to 
replace the implied assumption of a 1:1 relation between concentration (C) and EC that is 
inherent in Equation [37] with a more appropriate curvilinear relation. These modifications were 
calculated using a steady-state chemical model (Rhoades and Merrill, 1976; Oster and Rhoades, 
1990; Rhoades et al., 1992) that incorporates the effects of salt precipitation (S p ) and mineral 
weathering (S m ) reactions which occur in irrigated root zones on the resulting concentrations of 
solutes in the soil water and that calculates the corresponding EC value from these 
concentrations. The following curvilinear relation, which he obtained for the Colorado River 
water used for irrigation in the Wellton-Mohawk, illustrates the nature of the modification: 




1 

!5 20 25 



50 



35 



EC^ = 0.599 + 0.985 (^)- 0.008 (+)' 



[41] 



with an r value of 0.999 (see Figure 77). The value of EC iw was inherently incorporated into this 
relation by means of the coefficient values, as are the corrections for salt precipitation and 
dissolution and for conversion of total salt concentration to its' EC equivalent. These 
modifications permit EC to be used in place of the conservative chloride solute concentration and 
Equation [41] used in place of Equation [37] as a basis for estimating the steady-state value of 
leaching fraction. Additionally, he took steps to assure that the method would apply to each soil 
type encountered in the surveyed fields. To accomplish this, he established the slope and intercept 
values in Equation [5] relating EC a and EC, for each soil encountered in the project fields (see 



Soil salinity assessment 



89 



Figure 20, after Rhoades, 1980). He then 
substituted Equation [41] into Equation 
[5] with the assumption that EC WC = EC WS 
= EC dw (since the measurements applied to 
the bottom of the rootzone, where EC W = 
ECdw) to give specific equations relating 
EC a and E for each of the different soil 
types encountered in the surveyed fields. 
These equations took the following form: 



[y~) = slope EC a - intercept , 



[42] 



where the slope and intercept values were 
specific for each soil type (see Figure 78, 
after Rhoades, 1980). The appropriate- 
ness of the approach was tested in four 
different fields. The leaching fraction at 
each site was also estimated as the 
chloride-concentration ratio Cl iw /Cl sw , 
assuming the chloride ion was a 
conservative (tracer) solute that would 
behave ideally according to Equation [37]. 
The steady-state assumption was assumed 
applicable for the test-fields because the 
same crop had been irrigated for many 
years in each field with consistent 
management. This assumption was 
concluded to be appropriate, since 
essentially the same values of E were 
determined using both methods (Equation 
[42] and chloride ratio) of estimation (see 
Figure 79, after Rhoades, 1980). It was 
also concluded from these results that the 
more practical measurements of EC a could 
be used in place of the soil 
sampling/chloride ratio procedure to 
estimate leaching fractions. Since the 
composition of the drain water is defined 
in relation to either the leaching fraction 
value or EC a value in this approach, it is a 
simple matter to express the data in terms 
of absolute salt- load, provided V inf is 
known. This latter information, however, 
will be difficult to obtain in practical field 
appraisals, especially at the many different 
sites in the field where measurements of 
EC a are easily made. Analogously, the 
areal sources of salt-loading and the 
regions of high leaching can be readily 



FIGURE 78 

Calibrations established between leaching 
fraction (L f ) and electrical conductivity of soil 
(EC a ) for different soil types in the Wellton- 
Mohawk Irrigation and Drainage District of 
Arizona, USA (after Rhoades, 1980) 



S 20 




o I 



FIGURE 79 

Correlation found between electrical 
conductivity of soil water (EC) and the ratio 
of chloride concentration in soil water (Cl w ) 
below the rootzone to that in the Colorado 
River irrigation water for the Indio fine silty 
loam soil at four study sites in the Wellton- 
Mohawk Irrigation and Drainage District of 
Arizona, USA (after Rhoades, 1980) 



di 


^—^— i 


— r r 1 

Th«ctl!itSl 




20 




delation F::r *^ 








£»■<* 










5 I5 

■a 








iii 


* • 


if ^"~E*pwim*ii1<ll data 


■ 






r"*0,95 




ft 


AT 


ECw 064911*00(1/1-,) 


- 




'* 


-OOl5tl/Lrf 
i j. 1 





'0 5 10 13 20 23 

Ralio of Chtofid* Conc«Hno1imi in Sw* W^ler 
1* Colorado Ri war . I/L i 



90 Example uses of salinity assessment technology 



inferred and delineated by representing the data in the form of a leaching fraction map. Of course, 
inherent to the success of this approach are the following assumed conditions: (i) the soil is at 
steady-state, (ii) the chemical composition of the irrigation water is known and essentially 
constant over time, (iii) the soil water is near field-capacity at or below the lower extremity of its 
root zone, (iv) the relation between soil electrical conductivity and soil water electrical 
conductivity is known for the soil type that exists at that depth and field-site, and (v) there is no 
upward flux of water and salt from the shallow water table into the depth of EC a measurement. 
For cases where the latter condition exists, or other such non steady-state effects exist, an 
analogous approach based on Equation [36] or [38] could be developed to estimate leaching and 
salt- loading amounts. The advantage of this spatially-calibrated EC a -sensor approach is that a 
detailed pattern of leaching and salt-loading, and of their variability within the field, can be more 
practically obtained using the mobilized system of calibrated- EC a measurements described earlier 
than is possible using the soil sampling/laboratory procedures employed by the preceding 
investigators. Additionally, the salt-load can be more accurately estimated using the approach of 
Rhoades (1980), because the chemical composition of the drainwater is known as a function of E 
by means of the chemical model which is incorporated into the approach. 

The above example, as well as those of Cook et al. (1989) and Slavich and Yang (1990), 
studies have demonstrated that a generally good potential exists for the assessment of leaching 
rates in fields and for their associated salt-loads using relatively simple salt-balance models and 
geophysical sensors which measure soil electrical conductivity. This is especially true, if these 
models are modified by a method analogous to that described above so as to correct for some of 
the errors associated with the simplifying assumptions inherent within them. However, no one to 
date has combined all of these refinements together with the measurement/calibration methods 
into a practical system for such an assessment, especially one that permits the infiltration 
amounts to be established at the many points of measurement that are needed to account for the 
highly variable conditions of irrigation and leaching that typically exist in agricultural fields. 
Some data illustrating the degree of this variability and of the need to account for it in the desired 
assessments will be presented and discussed in the following paragraphs; additionally, some 
procedures will be suggested to make the assessments more practical than was possible when 
many of the previously discussed methods were developed. 

The practicality of the application of the salt-balance approach to the estimation of leaching 
and salt-loading rates has been markedly enhanced through the development by Rhoades and 
collaborators of the integrated, mobilized system of soil electrical conductivity measurement and 
of the salinity calibration software that was described earlier. This technology now makes it 
practical to obtain detailed accurate information of soil salinity distributions in irrigated root 
zones and fields. This technology has already been demonstrated to be capable of providing 
useful, qualitative information about irrigation uniformity, the adequacy of drainage, the relative 
degree of leaching, and the major source areas of excessive deep percolation and salt-loading 
(Rhoades et al., 1997a; 1997b). The utilization of this technology in combination with improved 
salt-balance models offers good potential to make rapid quantitative estimates of leaching and 
salt-loading rates. Some examples will now be given to support these conclusions and to illustrate 
the utility of this technology. 

Many of the examples already given in the section Evaluating adequacy and appropriateness 
of irrigation/drainage demonstrate how the relative degree of leaching can be inferred from the 
level and pattern of soil electrical conductivity within the root zones of irrigated fields. Figure 67 
illustrates the often observed relatively high leaching that occurs in the "upper" sections of 



Soil salinity assessment 91 



furrow-irrigated fields, as does Figure 68. With reference to Figures 70 and 71, it was shown 
how the net leaching fraction could be inferred from the mean salinity level of the root zone and 
chemical-model (WATSUIT) predictions assuming steady-state conditions. Additionally, these 
data and those of Figure 72, show how variable leaching can be from place to place within 
individual fields and how useful the mobilized systems of salinity assessment can be to determine 
this variability, its patterns and causes. With reference to Figures 74 and 75 and Tables 4 and 5, 
it was shown how the shape of the salinity profile could also be used to advantage for 
ascertaining and mapping those regions of the field where the relative leaching flux was 
excessive, or inadequate, as well as for evaluating the suitability of the irrigation practices. The 
absolute amount of leaching can be inferred from such data provided the amount of infiltration 
can be established for the various sites, and in turn the associated salt-load can be calculated 
from the EC dw , which can be determined from the assessment measurements by the methods 
previously explained/demonstrated. Practical methods are not available to measure this spatial 
variation in infiltration amount. Hence, the levels of soil salinity can be used to estimate the 
distribution of leaching rates that have occurred within the field from the overall field-average of 
infiltration amount and , in turn, the distribution of the associated salt-loading from the rootzone. 
Such methodology needs to be implemented to achieve management that is efficient in water use 
and protective of the environment. All of the measurement aspects needed to do this have been 
presented in this report. 



Scheduling and Controlling Irrigations 

Practical means of scheduling and controlling irrigations to conserve water while avoiding yield 
loss from salinity or water deficiency have not been given much attention. Presently utilized 
typical methods of scheduling irrigations are based on measurements of soil water status and/or 
predicted evapotranspiration. These latter methods are inadequate for saline soil conditions 
because they do not take into account salt (osmotic) effects on water availability, which also 
depends on soil water depletion allowed between irrigations. Furthermore, direct measurements of 
soil water depletion or matric potential can not be used to control the leaching fraction which is 
required to prevent excessive soil salinity accumulation. For saline water, irrigations must be 
scheduled before the total soil water potential (matric plus osmotic) drops below the level which 
permits the crop to extract water at a sufficient rate to sustain its physiological processes without 
excessive loss in yield. The crop's root system normally extracts progressively less water with 
increasing soil depth because rooting density decreases with depth and because available soil 
water decreases with depth as the salt concentration increases (Rhoades and Merrill, 1976). 
Therefore the frequency of irrigation should be determined by the total soil water potential in the 
upper rootzone where the rate of water depletion is greatest. On the other hand, the amount of 
water to apply depends on stage of plant development and the salt tolerance of the crop and, 
consequently, should be based on the status of the soil water at deeper depths. In early stages of 
plant development it is often desirable to irrigate to bring the soil to "field capacity" to the depth 
of present rooting or just beyond. Eventually, however, water must be applied to leach out some 
of the salts accumulating in the profile to prevent salt concentration from exceeding tolerable 
levels. Thus, the amount of water required is dictated by volume of soil reservoir in need of 
replenishment and level of soil salinity in the lower root zone. 



92 



Example uses of salinity assessment technology 



Since soil electrical con- 
ductivity is a tracer of the 
interactions of water infiltration, 
evapotranspiration and leaching as 
demonstrated above, it can be used 
as the basis for irrigation/salinity 
management. As shown earlier soil 
water salinity (and hence osmotic 
potential; M Pa at 25 C° = 0.04 x 
EC 2 5, in dS/m) and leaching 
fraction can be determined from 
measurements of soil electrical 
conductivity. Also as shown earlier, 
EC a is, for a given soil, mostly 
responsive to the EC (hence 
osmotic potential) of the soil water 
in the pores which supply most of 
the water to the plant and it can be 
used to determine leaching fraction. 
With calibration for the particular 
soil-type, the total soil water 
potential can be determined from 
EC a and @ w using Equation [5] and 
knowledge of the matric potential- 
@„ relation for the soil and the 
leaching fraction can be determined 
solely from EC a . The means for this 
and data showing its feasibility 
have been demonstrated by 
Rhoades efa/.(1981). 

Rhoades et al. (1981) showed/ 
concluded that the EC a measure- 
ments could be made in the upper 
profile to schedule irrigations and 
the latter measurements could be 
made in the lower profile to 
determine that sufficient, but not 
excess, water is being applied over 
the long-term to keep salinity within 
acceptable limits. As also shown by 
Rhoades et al. (1981), one can also 
associate a "set-point" value of EC a 
(equivalent to a desired total water 
potential) to use as a basis for 
scheduling irrigations, but the 
combined use of moisture and 
salinity measuring sensors would 
likely be more accurate (evidence of 



FIGURE 80 

Water penetration following a small irrigation, as 
deduced from measurements made with four- 
electrode sensors and a neutron probe (after 
Rhoades et al., 1981) 






i? 
J* 
*i 

.9 
J 

12 
■J 

& 
$ 
2 



jjH Pw*,« 



0» 




. !_ 




Timt , Jul4f1 Cfly* 



FIGURE 81 

Water penetration following an irrigation of 
moderate amount, as deduced from measurements 
of soil electrical conductivity (EC a ) with four- 
electrode sensors (after Rhoades et al, 1981) 




260 



B53 



2SS 
Tlmt, Julian Dayi 



;*.■■■> 



Soil salinity assessment 93 



this is shown in Figures 22 and 80). For such scheduling and monitoring purposes, the 
measurements of EC a can be made at monitoring locations using burial-type four-electrode 
sensors or estimated at many places in the field using the above-ground sensors and profile 
estimating procedures, as described earlier. 

At the end of an irrigation cycle, a certain EC a -depth relation will exist through the soil profile 
for any given combination of soil-, plant- and water- type. Upon irrigation, the water content will 
increase at every depth in the profile reached by the wetting front and EC a will increase 
correspondingly as water flows into that soil volume and @ w increases (see Figures 80 and 81). 
Thus, in principle, irrigations can be automated using EC a -sensors placed in the profile at desired 
depths to initiate irrigations when the set point value is reached and to terminate them when the 
EC a reading at the desired depth shows the arrival of water and/or sufficient leaching to keep 
salinity within limits. An example of the use of EC a -depth measurements to sense the movement 
of an irrigation wetting front is shown in Figure 81 . 

The theory and data presented here and in Rhoades, et al. (1981) support the conclusion that 
measurements of soil electrical conductivity could be used to schedule irrigations, control the 
depth of water penetration, and obtain the desired leaching fraction. Some irrigation systems 
could be automated with burial-type sensors. The methodology has good potential that should be 
exploited where salinity is a limiting factor. 



Reclamation of Saline Soils 

Though excessive levels of salts in soils can sometimes be reduced over time using management 
practices that are compatible with cropping, it is more usual to set aside cropping temporarily and 
to speed the removal process by reclamation practices. The selection of appropriate reclamation 
practices requires knowledge of the cause and source of the salt-related problems. Some examples 
were given in earlier sections to illustrate the utility of the salinity assessment methodology to 
help determine such causes and sources. 

The only practical way to reduce excessive soluble salts in soils is to leach the salts out by 
passage of lower-salinity water through the active rootzone depth of the soil. The amount of 
leaching required to reclaim saline soils is a function of the initial level of soil salinity, of the level 
desired and the depth of soil needed to be reclaimed (which are largely determined by the crops to 
be grown), and of certain soil and field properties and the method of water application (which 
influence leaching efficiency). Several theories have been developed to predict needed leaching 
but various required parameter "unknowns" usually limit their usefulness and accuracy without 
on-site calibration. For this reason empirical relations, which are based on field experiments or 
experience, are generally used as guidelines for reclamation. The accuracy of these guidelines are 
unknown for most situations. 

A simple, straight-forward way to determine the amount of leaching required for saline soil 
reclamation for a particular field and method of water application, or to follow its rate of 
accomplishment, is to initiate leaching of a test site by the intended method of water application 
and to follow the change in soil salinity with amount of water application/infiltration. A very 
convenient, practical monitoring approach is to measure changes in soil electrical conductivity, in 
this regard, using any of the sensor techniques previously described (though a burial-type four- 
electrode EC-probe would be more generally appropriate). The progress of salt removal is 
immediately evidenced from the EC a readings during leaching; this clearly and simply establishes 



94 



Example uses of salinity assessment technology 



the required leaching for that field and 
water application condition. The data 
shown in Figure 82 (after Rhoades, 
1979b) illustrate and demonstrate this 
method for a simple situation where a 
"salinity-probe" was installed at a 
depth of 15 cm for monitoring 
purposes. Analogous data would be 
used to monitor the deeper depths and 
rate of progress. 



The process of detecting the degree 
of change resulting during reclamation 
process is the same as that previously 
described for monitoring salinity; 
hence, it will not be repeated. The 
changes resulting from the reclama- 
tion processes in the mean value of 

salinity and in its spatial pattern for a given field would be determined by the same set of 
calculations provided in Annex 9. 



FIGURE 82 


Plot of soil electrical conductivity vs. depth of 
water infiltrated during the ponded leaching of 
salinized Pachappa soil (after Rhoades ,1979) 


7 

1 
V) 

™ g 




A 


m 

I.I 3 


\ 


y 
I* 


\ 


* 
if 

LU 1 


V^ 


H 


s ' st *— •— — . 




j i i ; i it. ii ■• * i 


Daplti ci LcacKIng, cm 



Soil salinity assessment 95 



Chapter 5 

Operational and equipment costs associated 
with salinity instrumentation measurement 

techniques 



Salinization is the increase in concentration of total dissolved solids in soil and water. Secondary 
salinization is the terminology commonly used for any type of human activity which increases the 
salinization of land or water resources. Secondary salinization of land resources has been 
occurring since the beginnings of human settlement, and is most often directly related to the 
development and expansion of irrigated agricultural practices. 

Secondary salinization reduces agricultural crop yields, degrades land values, and if left 
unabated, eventually leaves the affected soil in an unusable state. The economic costs of 
secondary salinization at both the farm and regional scale have been well documented. Grieve et 
al., explored the economic costs of waterlogging and soil salinity to the Murray Valley basin in 
New South Wales, Australia (Grieve et al., 1986). They estimated these costs to be in excess of 
16 million Australian dollars on an annual basis, which represented 16% of the district's total 
agricultural production. In a follow up study in 1994, Oliver et al., performed a capital cost 
survey of 177 local government agencies and 39 public utilities located in the greater Murray- 
Darling Basin (Oliver et al., 1996). They found that over 27.7 million Australian dollars were 
allocated and/or distributed for salinity related repairs and maintenance, and for salinity related 
research and education. In 1994, Luke and Shaw performed economic assessments of the costs of 
salinity to agriculture in the Loddon, Campase, and Avoca dryland sub-regions of Australia 
(Luke and Shaw, 1994a,b,c). They found that if left unabated, by the year 2001 the total 
economic losses for salinity in these three regions could reach 880, 323, and 531 thousand 
Australian dollars, respectively. 

The total area of salt-affected land on a global basis has been estimated to be 
approximately 76.3 million hectares (Mha), of which 41.5 Mha is considered to be seriously 
degraded (Oldeman et al., 1991). Serious secondary salinization is occurring at an ever 
increasing rate across the world's irrigated agriculture, and is responsible for substantial 
economic losses in agricultural production. Overall, the ratio of salt-affected to irrigated land has 
been estimated to be between 9 to 34% in the following countries: Argentina, 33.7%; Australia, 
8.7%; China, 15%; Commonwealth of Independent States, 18.1%; Egypt, 33%; India, 16.6%; 
Iran, 30%; Pakistan, 26.2%; South Africa, 8.9%; Thailand, 10%; and in the United States, 23% 
(Ghassemi, Jakeman, and Nix; 1995). 

Secondary salinization can be effectively controlled, provided proper land management and 
agricultural production methodologies are employed (Rhoades, 1997). However, the successful 



96 Operational and equipment costs 



implementation of agricultural salinity management strategies depends upon many political, 
economic, and technical factors which must be synthesised together in an objective fashion. In 
general, this process is not possible unless the magnitude and distribution of the soil salinization 
can first be quantified. Therefore, the inventorying and/or monitoring of soil salinity in a cost 
effective manner represents a critical first component to this management process. 

Within the last 25 years, a tremendous amount of progress has been made in regard to the 
assessment of soil salinity using electrical conductivity measurement techniques. These survey 
instrumentation techniques have been shown to be both highly accurate and rapidly employable, 
and in general represent the most cost effective salinization inventorying methodologies currently 
in use. However, to date these techniques have not been fully realized nor taken advantage of. 

In order to objectively quantify the cost benefits of the above instrumentation 
methodologies, an economic cost analysis must be performed. Such an analysis must include two 
components; (1) a detailed description of the capital costs associated with the various equipment 
used in the most common survey instrumentation techniques, and (2) an assessment of the 
operational costs incurred when applying these techniques for the measurement of soil salinity. 



Salinity instrumentation: equipment specifications and cost information 

Before an operational cost analysis can be performed on any type of soil salinity survey 
instrumentation technique, the performance specifications and capital costs of the instrument 
must first be determined. Hence, a summary of equipment specifications and product cost 
information for the most commonly used field salinity assessment instruments is given below. 
This discussion includes information concerning both manual (hand-held) instruments and 
mobilized salinity survey systems, global positioning systems (GPS), relevant analytical 
equipment, and soil salinity assessment and mapping software. 

Soil Salinity Survey Instruments 

Non-invasive Electromagnetic Induction Instruments 

EM38 

The Geonics EM38 was designed specifically for agricultural soil salinity surveys, and can be 
used to survey large areas quickly without employing ground (contact) electrodes. The EM38 
uses Geonics patented electromagnetic induction principle, providing depths of exploration of 1 .5 
meters and 0.75 meters in the vertical and horizontal dipole modes, respectively. 

The EM38 is very lightweight (2.5 kg), compact (1 meter long), and highly durable. It can 
be used to measure either apparent conductivity in millisiemens per meter (mS/m) or the inphase 
ratio of the secondary to primary magnetic field in parts per thousand (ppt). Either conductivity 
or inphase measurements can be collected in both the horizontal and vertical dipole modes. 
Measurements are normally made by placing this instrument on the ground and physically 
recording the meter reading. Signal readings can also be digitally logged by using a data logger in 
conjunction with the meter (see DL720 Data Logger, described below). Measurements can be 
made either manually (using a trigger switch) or in a continuous mode. Both the meter and data 
logging system can be operated by a single surveyor. 



Soil salinity assessment 97 



The EM38 is well suited for mechanized applications, since it can be easily mounted to 
various types of transport vehicles and/or towing platforms and its data acquisition software can 
be readily modified to interactively communicate with various types of computer systems, 
controller boards, and/or other survey instrumentation (such as GPS survey systems, etc.). 

DL720 Digital Data Logger 

The DL720 digital data acquisition system is a rugged and weatherproof logging system designed 
to support all Geonics ground conductivity meters. The DL720 system includes an Omnidata 
Polycorder 700 series data logger, interconnecting cables, and DAT software for data storage and 
manipulation. 

Conductivity data can be collected and stored in either static (manual) or continuous mode 
(using a time interval selected by the surveyor). Stored data are downloaded from the data logger 
directly into the DAT software program using RS232 serial connections. All Geonics DAT 
software has been designed for use on IBM compatible personal computers, and can be used to 
edit, plot, and print profiles of either the conductivity or inphase signal data. The data logging 
software also includes a time-stamp option, which facilitates the merging of any DAT data file 
with many common GPS file formats. 



Invasive (in situ) Electromagnetic Induction Instruments 

SCT-10 Conductivity Monitoring System 

The Martek Model SCT-10 is a portable, monitoring system capable of in situ measurement of 
soil conductivity and temperature. The SCT-10 is small, light weight, and contains its own 
rechargeable battery power supply, internal clock, memory, and communications port. All meter 
calibration and computations are performed via a low power, CMOS microprocessor while a 
large, alphanumeric display provides instructions and data in simple English. 

The SCT-10 accepts a wide assortment of conductivity sensors for soil and water 
applications and is capable of accepting all common cell constants and adjusting any collected 
conductivity data to 25 degrees centigrade. Signal data measurements are displayed in direct 
engineering units, and simultaneous digital and analog signals are available for secondary 
recording instruments. For recording data, the SCT-10 comes standard with 0-1 volt DC for 
analog recorders and a serial ASCII port for communication with computers or controlling 
devices. 

Four types of soil sensors are available for use with the SCT-10; including a (i) vertical 
sensor, (ii) horizontal array, (iii) bedding sensor, and (iv) burial sensor. The vertical sensor 
(commonly referred to as a standard soil-probe, insertion four-electrode probe, or Rhoades probe) 
can be used to acquire conductivity readings down the first 1.2 meters of the soil profile. The 
bedding probe is a miniature version of the vertical sensor, and is designed primarily for 
acquiring near-surface conductivity readings in seed beds or high density root areas. The burial 
sensor is an in situ sensor which can be left buried in the ground to facilitate the long term 
monitoring of soil conductivity over time. The last soil sensor, the horizontal surface array, can 
be used to determine the average conductivity of large volumes of soils by manually varying the 
configuration of the current electrodes in the array. 



98 



Operational and equipment costs 



Equipment Specifications and Product Cost Information 

Detailed equipment specifications and product cost information for the Geonics EM38 meter and 
DL720 digital data logger are listed in Table 6. Equipment specifications and cost information for 
the Martek SCT-10 meter and sensors are given in Table 7. 



TABLE 6 

Geonics EM38 and DL720 equipment specifications and costs. 



EM38 


Measurements 


conductivity (mS/m) or inphase ratio (ppt) 




Sensor 


dipole transmitter 




Intercoil spacing 


1 meter 




Operating frequency 


14.6 kHz 




Power supply 


9 volt alkaline battery 




Conductivity range 


100 to 1000 mS/m 




Inphase range 


±29 ppt 




Resolution 


±0.1% of full scale 




Accuracy 


±5% at 30 mS/m 




Instrument dimensions 


103 x 12 x 12.5 cm 




Case dimensions 


117x19x13 cm 




Instrument weight 


2.5 kg 




Shipping weight 


10 kg (including case) 




Cost 


US$7 195 




DL720 Specifications 


Storage capacity 


16 500 1 -channel records; 10 000 2-channel records 




A/D resolution 


16 bits 




Dimensions 


20 x 10 x 5.3 cm 




Weight 


1.5 kg 




Cost 


US$ 3 775 (includes DAT software and interconnecting 


cables) 



TABLE 7 

Martek SCT-10 equipment specifications and costs. 



Measurements 


temperature (°C) and conductivity (milli S/cm) 


Sensor types 


vertical, horizontal, bedding, in-plant (in situ), flow-through (liquid) 


Power supply 


rechargeable, internal nicad battery 


Conductivity range 


0-1, 0-10, 0-100 milli S/cm 


Temperature range 


0-50 °C 


Resolution 


±0.0001, ±0.001, ±0.01 milli S/cm; ±0.01 °C for temperature 


Accuracy 


±0.01 , ±0.05, ±0.5 milli S/cm; ±0.1 °C for temperature 


Instrument dimensions 


16.5 x 10.1 X20.3 cm 


Instrument weight 


2.7 kg (not including sensors) 


Cost (add 30% for foreign 


SCT-10 meter: US$2,000 


orders) 


Probes: (i) Vertical probe: US$500; (ii) Horizontal probe: US$300; 




(iii) Bedding probe: US$250 and (iv) Burial probe: US$100. 



GPS Equipment 

Other useful instruments for soil salinity survey work include, but are not limited to, soil 
temperature probes (for measuring soil temperature throughout the soil profile), GPS survey 
equipment (for acquiring spatial location), laser range finding systems (for measuring changes in 
micro-elevation), and various sensors designed to measure other types of soil or crop attributes 
(such as infrared spectrometers for measuring crop biomass or time domain reflectometric 
sensors for measuring volumetric soil water content). Amongst all these instruments, a reliable 
and versatile GPS system is by far the most important. 

There are three primary reasons why a good GPS system should be employed when 
performing soil salinity survey work. First, salinity survey data are almost always spatial in 



Soil salinity assessment 99 



nature; i.e., survey data is generally acquired across a spatial region and hence the spatial 
locations of the survey sites must be known in order to construct any sort of contour or relief 
map. While it is possible to physically grid (i.e., measure) out these survey locations, this data 
can typically be acquired faster and more efficiently using a GPS system. Second, nearly all types 
of mechanized conductivity transport systems collect their signal data "on-the-go"; i.e., the 
conductivity data is acquired while the vehicle is continuously moving. This in turn requires the 
use of a GPS system, since this is the only type of commercially available system which can 
readily collect and record the continuously changing spatial location of a moving object (i.e., the 
moving transport vehicle). And third, nearly all commercial GPS units are designed to interface 
with multiple types of geographic information systems (GIS). Thus, the transfer of the spatial 
location information into a GIS is readily facilitated using GPS equipment. 

Typical GPS systems range in price from a few hundred to many thousands of dollars or 
more, depending on the system accuracy specifications. Most types of salinity survey work 
require survey location accuracies of ±1 to 2 metres, which can be obtained from GPS units 
which facilitate differential correction (the typical cost of a differentially correctable GPS unit 
starts around US$1000). Differential correction can be obtained in two ways; either through post- 
processing or real-time. Post processing (i.e., correcting the data after it has been collected and 
downloading from the GPS unit) is usually more accurate and less expensive. However, real-time 
differential correction is generally required when one must navigate to pre-determined survey 
locations. 

The largest commercial GPS companies and satellite differential correction suppliers 
include Trimble Navigation, Magellan, Garmin, Ashtech, Racal, and Ominstar. These companies 
can be contacted directly for current product line and pricing information. 

Mobilized Soil Salinity Survey Systems 

Mobilized Systems for Non-invasive Instruments (EM38) 

In its simplest form, the mobilization of the EM38 can be achieved by simply mounting the 
instrument on some type of sled or trailer and then towing it over the survey area. The sled or 
trailer must be completely non-metallic and must be kept at least 1 to 2 meters from the towing 
vehicle (which can be either a small tractor or an alternative terrain vehicle). A GPS system must 
also be set up and mounted on the towing vehicle (or more preferably, to the trailer itself) in order 
to simultaneously record spatial location information. The EM38 would be set to record 
conductivity data in a continuous mode (defined by a time interval which may be selected by the 
surveyor) and this data would then be automatically "time-stamped". Hence, once the survey was 
completed, this data could be readily merged with most types of post-processed GPS location 
information to create a spatially referenced conductivity map. 

There are currently no third-party vendors who sell commercially available, "off-the-shelf 
sleds or trailers for the EM38. However, these types of towing apparatuses are fairly simple to 
fabricate, and can generally be built for between US$500 to US$3 000 by many trailer 
fabrication shops. 

More complex trailering platforms generally require data logger software modification 
and/or customized system integration. For example, a simple EM38 trailer design will typically 
support (i.e., physically hold) only one EM38 unit at a fixed height and orientation above the soil 
surface. To acquire both horizontal and vertical EM38 signal data, the trailer must support and 
control two EM38 units (operated in sequence), or the trailer must be able to mechanically rotate 
the unit (as well as control the timing of the data logging). Additionally, it is usually desirable to 



1 00 Operational and equipment costs 



directly interface the GPS system with the DL720 data logger so that either system can 
communicate (and/or control) the other. Again, there are currently no commercially available 
systems which are designed to perform these sort of survey operations. However, there are 
geophysical companies which can perform the software modification and/or system integration 
necessary to create such trailering platforms. One such company with previous commercial 
experience in Geonics DL720 custom software modification and EM38/GPS system integration 
is Geomar Geophysics Ltd., in Toronto, Canada. Geonics Limited can also sometimes 
recommend one or more appropriate companies for the design, integration, and fabrication of 
advanced trailering platforms. 

The total cost of such a trailer obviously depends on the mechanical and electronic 
complexity of the platform design. Hence, estimating an approximate cost for such a trailering 
system is impossible without first knowing the design specifications. However, a very gross price 
range for most types of complex platforms would be between US$2 500 to US$10 000. (Note 
that this cost does not include the EM38, GPS system, or towing vehicle.) 

An even more sophisticated and versatile approach to mobilized surveying can be 
employed by suspending and controlling the EM instrument(s) directly from a specialized, self 
contained transport vehicle. This sort of system has been developed at the United States Salinity 
Laboratory by Rhoades and colleagues (Rhoades 1992a, 1992b, 1993, 1994, 1996b; Carter, et 
al., 1993) and is currently being developed for commercial distribution by Agricultural Industrial 
Manufacturing (AIM) Inc., in Lodi, California. This automated transport system incorporates 
both the use of a Geonics EM38 meter and modified Martek SCT-10 meter / horizontal array 
system, along with a synchronized GPS system for recording spatial location information. This 
automated transport system can also be readily adapted to incorporate additional sensors. Retail 
pricing information for this system is not yet available; however, Table 8 lists the price 
breakdown of the various costs associated with fabricating the basic hydraulic transport vehicle 
(available from West Texas Lee Company Inc.) into a multi-purpose salinity assessment vehicle. 

TABLE 8 

Price breakdown for fabricating a basic hydraulic transport vehicle into a multi-purpose salinity 

assessment vehicle. 



Design and fabrication category 


Modification costs, 






us$ 


1 . Construction of front mast assembly with EM38 tube and rotating unit 




5 883 


2. Horizontal array probe assemblies and scissors frame structure 




5 900 


3. Electronic controls, computerized control system and programming 




6 450 


design 






4. Wiring, limit switches, enclosures, and control box 




2 680 


5. Hydraulic control valves, manifold, and plumbing 




2 560 


6. Console, mounting brackets, and accessories 




1 490 


7. Parking brakes on drive wheels, positive neutral on hydrostatic pump 




1 980 


8. Miscellaneous: bracket, paint, etc. 




990 


Total modification cost 




27 933 


Note: total modification cost does not include the costs of the EM38, DL720, 


SCT-10, 


GPS, or cost of 


the basic hydraulic transport vehicle. 







Mobilized Systems for Invasive Instruments 

Like the non-invasive EM instruments, mobilized versions of conductivity sensors using four- 
electrode technology have been developed for both research and commercial applications. The 
most common systems employ one or more sets of four electrodes (in the form of either 



Soil salinity assessment 101 



penetrating shanks or circular disk blades) which are mounted to either a tractor tool bar or trailer 
platform. This tool bar or trailer platform is then towed or dragged across the survey area, 
allowing for the near continuous collection of conductivity data (up to one reading per second). 
These systems are also typically designed to interface with and/or incorporate simultaneously 
recorded GPS location information. 

A mobilized, tractor-mounted version of the horizontal fixed-array conductivity sensor 
system has been developed at the United States Salinity Laboratory by Rhoades and colleagues 
(Rhoades 1992a, 1992b, 1993, 1994, 1996b; Carter et al, 1993) and is also currently being 
developed for commercial distribution by AIM Inc., in Lodi, California. Retail pricing 
information for this system is not yet available; however, Table 9 lists a price breakdown for 
assembling such a sensor system. A commercial version of a similar four-electrode system 
(employing the use of circular disk blade technology) is currently available from Veris 
Technologies in Salina, Kansas. This latter system simultaneously collects two sets conductivity 
data by employing two sets of horizontal four-electrode arrays, and can be towed using an 
alternative terrain vehicle (ATV) or standard tractor. However, this system is not designed to be 
operated in a bed-furrow environment (the spacings between the disk blades are not adjustable 
and the clearance ratio between the toolbar and the soil surface is less than 30 cm). More detailed 
equipment specifications and product cost information for the Veris 3100 System are given in 
Table 10. 

TABLE 9 

Estimated system equipment costs for building a mobilized, horizontal array conductivity sensor 

system 



Component 






Estimated costs, 
US$ 


1. SCT-10unit 






2 000 


2. SCT-10 modifications 






1 000 


(includes pre-amp, RS232 serial interfacing, and circuit board 








modifications) 








3. Tractor mounted tool-bar with adjustable vertical penetrating sh 


anks 




2 000 


4. Wiring and instrument control box 






250 


Total system cost 






5 250 


Note: total estimated system cost does not include the cost of the 


GPS 


system 


or tractor. 



TABLE 10 

Verris 3100 system equipment specifications and costs 



Item Cost, US$ 

Verris 3100 Soil mapping system, with standard features: 1 1 000 

- all-welded tubular steel frame 

- heavy duty spring-loaded coulter/electrodes 

- non-invasive 17 inch flat disk blades 

- P205 R75 highway tires 

- ratchet raising/lowering system 

- adjustable 4-position clevis hitch 

- built in micro-processor 

- 3.5 inch floppy disk drive 

- internal flash memory to store up to 10 hours of data 

- back-lit transflective display 

- DGPS compatible RS232 serial port 

Road kit: ball hitch, lights 175 

Weight package 725 

Notes: (1) Conductivity measurements acquired in milli S/m; supplier should be contacted directly for 

conductivity range, resolution, and accuracy specifications under various operating environments. 

(2) Verris 3100 system cost does not include GPS or towing vehicle costs 



1 02 Operational and equipment costs 



Analytical (Laboratory) Conductivity Instruments 

There are numerous analytical conductivity meters which are capable of measuring the 
conductivity of a solution extract. The cost of such equipment can be quite variable, depending 
upon the specific instrument specifications. However, there is one conductivity meter which 
deserves special mention because of its unique ability to measure soil conductivity in a saturated 
soil paste; the Hach CO150 Conductivity Meter. 

The basic Hach CO150 system includes a portable battery operated meter, one 
conductivity probe for solution extracts, and one soil cup for measuring conductivity of a 
saturated soil paste. The CO 150 system also includes all the necessary software to estimate soil 
salinity from the saturated paste conductivity reading, based on the methodology of Rhoades et 
al., 1989b,c. This latter ability makes the CO150 system uniquely different from all other 
commercially available conductivity meters, since it is the only meter to incorporate this 
technology. The basic CO150 system is scheduled for commercial release in early 1998. (An 
upgraded system is also scheduled for release in 1998 which will include a sodium probe, pH 
probe, and additional software for estimating the sodium adsorption ratio from the saturated paste 
conductivity, sodium, and pH readings.) Advanced equipment specifications and retail price 
information for the basic conductivity system are given in Table 11. 

TABLE 11 

Equipment and estimated cost specifications for basic Hach CO150 conductivity meter 



Measurements 


temperature (°C), conductivity (milli S/cm) and total dissolved solids (mg/L) 


Electrode types 


conductivity probe (liquid), soils cup (saturated paste) 


Power supply 


9 volt alkaline battery ( 9 VDC line adapter to 1 15 or 230 volts also 




available) 


Conductivity range 


0-0.2, 0.2-2, 2-20, 20-200 milli S/cm 


Temperature range 


-10-110 °C 


TDS range 


Oto 19 900 mg/L 


Resolution 


3 significant digits in conductivity or TDS mode 




±0.1 °C for temperature 


Accuracy 


±0.5% of full scale within each range of conductivity 




±1 .0 °C for temperature 




±1% RSD, 5 to 70 °C for TDS 


Instrument dimensions 


20.5 x 8.3 x 4.8 cm (not including probe or soils cup) 


Data logging 


50 data set storage; LCD display output and RS232 output 


Cost 


US$1 295 to US$1 895(estimated, includes one probe, soils cup, and 




software) 



There can be significant cost savings associated with the measurement of soil salinity in the 
saturated soil paste, as opposed to a solution extract. This is due primarily to the speed in which 
samples can be processed and the ability to forego the use of a vacuum extraction system. 
Additionally, saturated soil paste conductivity measurement systems make it simpler and more 
inexpensive to process salinity samples "in house", and thus reduce (or eliminate) the need to 
send the soil samples out for commercial laboratory analysis. The various costs associated with 
commercial versus internal laboratory sample analyses are discussed in detail in the section 
Operational costs associated with the appraisal of soil salinity. 

Soil Salinity Assessment and Mapping Software 

All instrumental salinity assessment methods used for field survey work require that measured 
conductivity readings be somehow converted into (estimated) salinity levels. This conversion can 



Soil salinity assessment 1 03 



be made using either a deterministic (theoretical) or stochastic (statistical) model. In either case, 
computer programs can be employed to perform the necessary conductivity- to-salinity 
calculations, and/or generate the estimated salinity map. 

All of the formulas needed to estimate soil salinity from bulk soil conductivity data in 
conjunction with known or estimated secondary soil properties is given in Rhoades et al., 
(1989a). Statistical software for estimating spatial regression conductivity-to-salinity models 
(given a limited set of calibration soil samples) is available from the United States Salinity 
Laboratory (the ESAP Software Package, version 1.0; Lesch et al., 1995c). ESAP is public 
domain software, and available free of charge. 

Numerous high quality contouring programs are available commercially. Any professional 
grade mapping and contouring package can be employed for the purpose of creating two or three 
dimensional conductivity and/or salinity maps. Retail prices for such software typically range 
between US$250 to US$750. 

Company Information 

Full addresses are given in Table 12 for all the salinity instrumentation companies discussed in 
this report. 



Operational costs associated with the appraisal of soil salinity 

To appreciate the potential cost savings which can be realized from employing survey 
instrumentation techniques, the overall operational costs for these various methods must be first 
determined and then compared to the costs associated with conventional sampling. Such an 
analysis will invariably depend on a number of economic and/or technical assumptions, and these 
assumptions should be clearly stated and explained. However, there are two underlying 
assumptions made throughout the entire cost assessment process which deserve to be elaborated 
on beforehand. 

First, it is important to realize that all of the various field instrumentation used for salinity 
survey work actually measure soil conductivity, and not soil salinity per se. Furthermore, the 
conversion from conductivity to salinity requires either (1) the accurate measurement or 
estimation of additional soil properties at each and every conductivity survey site, or (2) the 
collection of a limited set of "calibration" soil samples from the survey area under study. The 
latter approach is usually more cost effective, since acquiring soil samples at a few survey sites 
tends to be cheaper than acquiring soil property information at every survey site. (This approach 
is often referred to as "stochastic calibration", since it is based on statistical and/or geostatistical 
modelling techniques.) Therefore, throughout this discussion it is assumed that all but one of the 
survey instrumentation techniques require the collection of at least a few calibration soil samples, 
in addition to the collection of the survey data. The one exception is the insertion four-electrode 
survey, since this type of survey directly facilitates the estimation of the above mentioned 
secondary soil properties at each and every survey site. 

Second, there are two methods which one could use to describe the average cost of a 
survey instrument over its expected life time. The first would be on a cost per reading basis, and 
the second would be on a cost per time basis. Most of the instruments typically used for field 
conductivity surveys (such as the Geonics EM38 or Martek SCT-10 meter) go immediately into 



104 



Operational and equipment costs 



TABLE 12 

Addresses for salinity instrumentation companies 



Product 


Company 


Mechanized Salinity 
Assessment Vehicle 


Agricultural Industrial Manufacturing, Inc. P. O. Box 53 Lodi, 
California 95241 USA; Telephone: 209-369-1994 


Veris Technologies 

601 N. Broadway 
Salina, Kansas USA 67401 
Telephone: 913-825-1978 
Fax:913-825-2097 
web: www.veristech.com 


Custom GPS / EM38 Integration 
and / or Software Interfacing 


Geomar Geophysics, Ltd; Attn: Jerzy Pawlowski 

2-3415 Dixie Road, Suite 348; Mississauga, Ontario Canada L4Y 

4J6 

Telephone: 905-306-9215; fax: 905-276-8158 

e-mail :jerzy@geomar. com 


EM38 


Geonics Limited 

1745 Meyerside Drive, Unit 8 Mississauga, Ontario Canada L5T 

1C6 

Telephone: 905-670-9580 

fax: 905-670-9204 

web: www.geonics.com 


Conductivity Meter (CO150), Soil 
Cup System and Salinity and 
Sodicity Kits 


Hach Company 

P. O. Box 389 

Loveland, Colorado USA 80539 

Telephone: 800-227-4224 (in USA only), otherwise 970-669-3050 

fax: 970-669-2932 

web: www.hach.com 


Four electrode sensor, SCT-10 


Martek Instruments, Inc. 

2609 Discovery Drive, Suite 125 
Raleigh, North Carolina 27616, USA 
Telephone: 800-628-8834 
web:www. 4martek.com 


Eijkelkamp Agrisearch Equipment 

P O Box 4 

6987 Z G Giesbeek, The Netherlands 

Telephone: +31 313631941 

Fax: +31 313 632167 

web: www.eijkelkamp.com 


Elico Limited 

B-17, Sanathnagar Industrial Estate 
Hyderabad, 500 018, India 
Telephone: 040-22-2221 
Fax: 040-31-9840 


Soil Temperature Equipment 


Wahl Instruments, Inc. 

5750 Hannum Avenue 
Culver City, California USA 
Telephone: 310-641-6931 
Fax:310-670-4408 


TDR Equipment 


Environmental Sensors Inc. 

100-4243 Glanford Avenue 
Victoria, BC, Canada V8Z 4B9 
Telephone: 250-479-6588 
Fax:250-479-1412 



Soil salinity assessment 1 05 



a "scan" mode the moment they are turned on. Thus, these meters are continuously analysing 
signal data and calculating conductivity readings, regardless of whether the surveyor actually 
records a reading once per second or once per hour. Therefore, it seems appropriate to use the 
cost per time unit method for estimating survey equipment expenses. 

Operational Costs Associated with Conventional Sampling 

In order to facilitate a reasonable cost analysis of the various instrumental approaches, it is first 
necessary to establish the overall cost associated with a conventional sampling approach. In this 
context, "conventional sampling" means that all salinity information is acquired through the 
direct laboratory analysis of soil samples, without benefit of any type of secondary instrument 
information. 

Table 13 documents the various costs associated with such an approach for a typical 64- 
hectare field. The following assumptions have been made in the analysis shown in Table 13. 
First, the sampling is performed on a 12 by 12 grid, yielding 144 sample sites, and each site is 
sampled at 3 depths (0-30 cm, 30-60 cm, and 60-90 cm). Second, this process (the collection of 
three samples at one site) takes a two-person crew 15 minutes to complete, and this crew can 
walk across the field at a rate of 4 km per hour and lay out the survey grid as they walk. Third, 
the total walking distance is equal to 2 plus the number of transects times the physical length of 
the transect; note that the factoring in of 2 extra transects is done to cover the additional travel 
distance in between transects and the return distance (back to the starting point) after leaving the 
last sample site. Fourth, the total cost associated with each sample on a per sample basis is 
US$20.00, which includes the labelling and packaging in the field (US$0.50), handling and 
shipping costs (US$1.50), and external laboratory analysis costs (US$18.00). Fifth, after the soil 
salinity measurements are returned from the external laboratory, it will take 4 hours to input and 
process this data and generate a field salinity map. And finally, the two-person sampling crew is 
comprised of two technical support personnel paid at a rate of US$8.00 each per hour, and the 
data processing is performed by one technical specialist paid at a rate of US$20/hour. 

TABLE 13 

Cost analysis for a salinity appraisal of a typical 64-hectare field using a conventional soil 

sampling approach 



Assumptions 


800 m by 800 m field size (64 ha); 12 by 12 grid (144 sample sites); 3 




sample depths per site (0-30, 30-60, 60-90 cm depths) 


Time: 


Sampling : 5 minutes per 30 cm sample increment; Walking speed : 4 




km/hour 


Sampling costs: 


Labelling / packaging US$0.50/sample 




Shipping / handling US$1 .50/sample 




Laboratory analysis US$18/sample 




Total cost US$20/sample 


Total walking distance 


(number of transects +2) x (transect length) = 1 4 x (0.8 km) = 1 1 .2 km 


Total walking time 


11.2/4.0 = 2.8 hours 


Total sampling time 


(5 minutes / sample ) x (3 sample depths) x (144 sites) / 60 = 36 hours 


Data processing time 


4 hours 


Labour costs 


Specialist at US$20/hour 




Technical support at US$8/hour 


Crew size 


2 Tech Support for Sampling, 1 Specialist for data processing 


Total labour costs 


38.8 hours @ US$16/hour + 4 hours @ US$20/hour =US$700.80 


Total sampling costs 


(144 cores) x (3 samples / core) @ US$20/sample = US$8 640 


Overall Survey Cost 


US$9 340.80 per field (64-hectare) or US$145.95/ha 



1 06 Operational and equipment costs 



The total labour costs associated with this survey would be US$700.80 and the total 
sampling costs would be US$8 640. Hence, the overall survey cost associated with the 
conventional sampling approach would be US$9 340.80 per field (64-hectare) or US$145. 95/ha. 

From a farming for profit perspective, an operational cost of approximately US$146.00/ha 
is inordinately expensive, and hence non-justifiable. Furthermore, as Table 13 shows, the vast 
majority of this cost is incurred from the commercial laboratory analysis of so many soil samples. 
Therefore, any techniques which can be used to reduce the total soil analysis cost will also clearly 
generate the greatest financial savings (with respect to the overall survey cost). 

Obviously, one way to lower the analytical costs is to simply collect less samples. Of 
course, this is the primary idea behind each of the various instrument appraisal methods; i.e., to 
use the instrument readings as surrogate information (in place of actual soil samples). However, 
another way to achieve significant savings in the overall analytical cost is to forego the use of an 
commercial laboratory, and instead use some practical form of internal salinity appraisal method. 
In other words, one would set up their own laboratory and perform their own salinity analyses. 

The simplest (and most cost effective) way to run an internal "laboratory" for soil salinity 
determination is to use the saturated paste salinity appraisal methods described in Rhoades, 
1989b,c. In practice, such a laboratory would actually be nothing more than a work area having a 
sink, portable balance, a few basics laboratory supplies, and a conductivity instrument capable of 
measuring the conductivity of a saturated soil paste. 

Table 14 documents the cost savings which can be achieved by employing such an 
approach in conjunction with the Hach CO150 Conductivity System. The following assumptions 
have been made in the analysis shown in Table 14. First, the cost of the CO150 system is 
assumed to be US$1,500 and its expected lifespan is 15,000 samples, yielding an equipment cost 
of US$0.10 per sample. Second, the average time to make a saturated soil paste is assumed to be 
6 minutes, the average time to operate the meter and measure the paste is assumed to be 2 
minutes, and that all other additional laboratory costs incurred during this process amount to 
US$0.25 per sample, and third, one technical support person is used to make the soil paste, and 
one technical specialist is employed to operate the conductivity meter. Using these assumptions, 
the total labour cost incurred during the measurement of one sample would be US$1.46 and the 
total equipment cost would be US$0.35. This implies that the total salinity appraisal cost would 
be US$2.31 per sample (after adding in the US$0.50 per sample field labelling and packaging 
costs), which represents a 88% reduction in the laboratory analysis cost (down from 
US$20/sample). Hence, the recalculated cost of the conventional sampling method becomes 
US$1 698.72 per field (64-hectare) or US$26.54/ha. 

The cost savings calculated above represent the typical savings one would achieve by using 
the saturated soil paste appraisal method. Furthermore, the external laboratory reference cost of 
US$18/sample is conservative compared to most current US commercial laboratory rates (and 
hence the actual savings could be greater). None the less, these figures demonstrate that the 
CO 150 conductivity system would pay for itself after only one field survey. 



Soil salinity assessment 1 07 



TABLE 14 

Cost analysis associated with the saturated paste salinity appraisal method. 



Cost of Hach CO150 System 


US$1 500 


Expected life span 


15 000 samples 


Cost per sample 


US$0.10 /sample 


Other laboratory costs 


US$0.25/sample 


Average time to make saturated paste 


6 minutes 


Average time to measure saturated 


2 minutes 


paste 




Laboratory crew size 


1 Technical support for making paste, 1 Specialist to 




operate equipment 


Labour cost per sample 


0.10 hours/sample® US$8.00/hour + 




0.033 hours/sample® US$20/hour = US$1.46/sample 


Lab cost per sample 


US$0.10 per sample (CO150) + US$0.25 per sample 




(supplies) 


Revised sampling costs: 


Labelling/packaging US$0.50/sample 




Shipping/handling eliminated 




Laboratory analysis US$1.81/sample 




Total cost US$2.31 /sample 


Recalculated Cost of Conventional Sampling Method 


Total labour costs 


38.8 hours @ US$16/hour + 4 hours @ US$20/hour = 




US$700.80 


Total sampling costs 


(144 cores) x (3 samples/core) @ US$2.31 /sample 




US$997.92 


Revised Overall Survey Cost 


US$1 698.72 per field (64-hectare) or US$26.54/ha 



In general, significant cost savings can be realized using the saturated soil paste appraisal 
method. Furthermore, when this approach is used in conjunction with the various instrumentation 
methods, it becomes possible to achieve a substantial reduction in the total cost of a typical 
salinity survey. Therefore, although the remainder of this document will focus primarily on the 
various instrument appraisal costs, one should keep in mind that the greatest cost savings are 
always achieved through a combination of these two approaches. 

Operational Costs Associated with Survey Instrumentation 

As previously stated, the basic idea behind the various instrument appraisal methods is to exploit 
the instrument readings as surrogate information in place of actual soil samples. In principle, soil 
salinity can be calculated from soil conductivity provided the following additional soil 
information is known (or accurately estimated): temperature, saturation percentage, volumetric 
soil water content, and bulk density (Rhoades, 1989a). However, this additional soil information 
is usually not acquired during most surveys (with the exception of insertion four-electrode 
surveys). Hence, it usually becomes necessary to acquire a limited set of additional soil samples. 
These soil samples are typically referred to as "calibration" samples, because they are used to 
calibrate the soil conductivity to soil salinity through various statistical modelling techniques 
(such as regression or geostatistical models, etc.). 

A formal review of the various stochastic calibration techniques is beyond the scope of this 
discussion (the interested reader should refer to the main text of this book or Lesch et ai, 
1995a,b). However, the following brief comments are in order. First, an insertion four-electrode 
conductivity reading can generally be used to supply a more accurate estimate of soil salinity then 
a non-invasive (EM or horizontal array) reading. This is true because (1) the insertion four- 
electrode reading is depth specific, and (2) to acquire this reading one must first bore a hole into 
the soil, which in turn means that the soil removed from the bore hole is available for physical 
inspection (and hence the above mentioned secondary soil properties can be inferred from this 



1 08 Operational and equipment costs 



inspection). Second, in most field survey applications an ordinary regression model can be used 
for purposes of calibration. And third, any type of instrument survey will generally require 
calibration soil samples in order to achieve maximum prediction accuracy when accurate 
knowledge of the secondary soil physical properties is unavailable. Therefore, in the analysis 
which follows, the instrumental salinity survey costs have been broken down into two separate 
components: (1) the costs associated with acquiring the actual instrument survey data, and (2) the 
costs associated with acquiring the calibration soil samples. 

Table 15 documents the costs associated with acquiring EM38 survey data in a typical 64 
hectare field. The following assumptions have been made in the analysis shown in Table 15. 
First, the EM38 survey is performed on a 12 by 12 grid (yielding 144 survey sites), two EM38 
readings are acquired each site, and the data acquisition at each site takes 15 seconds. Second, the 
total walking distance and walking time is assumed to be the same as the distance and time 
required in the conventional sampling approach. Third, one hour is needed to perform all the post- 
survey EM38 data processing, and both the survey and data processing can be performed by one 
technical specialist (at a rate US$20/hour), and fourth, the total equipment cost is US$10 970, 
the expected equipment lifespan is 4 000 hours, and hence the average equipment cost can be 
estimated to be US$2.74 per hour. 

The total labour costs associated with this survey would be US$88 and the total equipment 
costs would be US$9.32. Hence, the overall cost associated with the EM38 survey would be 
US$97.32 per field (64-hectare) or US$1.52/ha. 

TABLE 15 

Operational survey costs associated with an EM38 survey in a typical 64-hectare field 



Assumptions: 


800 m by 800 m field size (64 ha); 12 by 12 grid (144 survey 




sites); 2 survey readings per site (horizontal and vertical) 


Equipment cost: 


US$7 195 (EM38) + US$3 775 (DL720) = US$10 970 


Expected life span: 


4 000 hours 


Cost per hour: 


US$10 970/4 000 = US$2.74/hour 


Time: 


EM38 survey readings (15 seconds/site); Walking speed: 4 km / 

hour) 

(number of transects +2) x (transect length) 1 4 x (0.8 km) =11.2 

km 

11.2/4.0 = 2.8 hours 


Total walking distance 


Total walking time 


Total survey time 


(0.25 min ) x (144 sites) / 60 min = 0.6 hours 


Data processing time 


1 hour 


Labour costs: 


Specialist US$20/hour 


Crew size 


1 Specialist for EM38 survey work and data processing 


Total labour costs 


4.4 hours @ US$20/hour = US$88 


Total equipment costs 


3.4 hours @ US$2.74 per hour = US$9.32 


Overall survey cost 


US$97.32 per field (64-hectare) or US$1.52/ha 



Table 16 documents the costs associated with acquiring SCT-10 insertion four-electrode 
survey data in the same 64-hectare field. The following assumptions have been made in the 
analysis shown in Table 16. First, the insertion four-electrode survey is performed on a 12 by 12 
grid (yielding 144 survey sites), three insertion four-electrodes are acquired each site (at depths of 
15, 45, and 75 cm), estimates of the secondary soil physical properties are acquired during this 
process, and all of the data acquisition at each site takes 4.5 minutes. Second, the total walking 
distance and walking time is assumed to be the same as the distance and time required in the 
conventional sampling approach. Third, one hour is needed to perform all the post-survey SCT- 
10 data processing, and both the survey and data processing can be performed by one technical 
specialist (at a rate US$20/hour). And forth, the total equipment cost is US$2 500, the expected 



Soil salinity assessment 



109 



equipment lifespan is 4,000 hours, and hence the average equipment cost can be estimated to be 
US$0.63/hour. 



TABLE 16 

Operational survey costs associated with an insertion four electrode survey (using a Martek SCT- 

10 meter) in a typical 64- hectare field 



Assumptions 

Equipment cost 
Expected life span 
Cost per hour 
Time 



Total walking distance 
Total walking time 
Total survey time 
Data processing time 
Labour costs 
Crew size 
Total labour costs 
Total equipment costs 
Overall survey cost 



800 m by 800 m field size (64 ha); 12 by 12 grid (144 survey sites) 

3 survey readings per site (0.15, 0.45, and 0.75 cm depths) 
US$2 000 (SCT-10 meter) + US$500 (Probe) = US$2 500 

4 000 hours 

US$2 500/4,000 hours = US$0.63 per hour 

SCT-10 Insertion 4-probe survey readings (1.5 min per 30 cm depth 
increment) 

Walking (4 km / hour) 

(number of transects +2) x (transect length) = 1 4 x (0.8 km) = 1 1 .2 km 
1 1 .2 km/(4 hours/km) = 2.8 hours 

(1.5 min ) x (3 readings per site) x (144 sites)/60 min = 10.8 hours 
1 hour 

Specialist at US$20/hour 

1 Specialist for SCT-10 survey work and data processing 
14.6 hours @ US$20/hour = US$292 
13.6 hours @ US$0.63 per hour = US$8.57 
US$300.57 per field (64-hectare) or US$4.70/ha 



The total labour costs associated with this survey would be US$292.00 and the total 
equipment costs would be US$8.57. Hence, the overall cost associated with the SCT-10 insertion 
four-electrode survey would be US$300.57 per field (64-hectare) or US$4.70/ha. 



Table 17 

Operational survey costs associated with a fixed, horizontal array survey (using a Martek SCT-10 

meter) in a typical 64-hectare field 



Assumptions 


800 m by 800 m field size (64 ha); 12 by 12 grid (144 survey sites) 




2 survey readings per site (1 m and 2 m span) 


Equipment cost 


US$2 000 (SCT-10 meter) + US$300 (Probe) = US$2 300.00 


Expected life span 


4 000 hours 


Cost per hour 


US$2 300/4, OOOhours = US$0.58 per hour 


Time 


SCT-10 Horizontal Array Survey Readings 




(1 minute/site); Walking speed: 4 km/hour) 


Total walking distance 


(number of transects +2) x (transect length) = 1 4 x (0.8 km) = 1 1 .2 km 


Total walking time 


11.2/4.0 = 2.8 hours 


Total survey time 


(1 min per site) x (144 sites)/60 min = 2.4 hours 


Data processing time 


1 hour 


Labour costs 


Specialist at US$20/hour 


Crew size 


1 Specialist for SCT-10 survey work and data processing 


Total labour costs 


6.2 hours @ US$20/hour = US$124 


Total equipment costs 


5.2 hours @ US$0.58 per hour = US$3.02 


Overall survey cost 


US$127.02 per field (64-hectare) or US$1.98/ha 



Table 17 documents the costs associated with acquiring SCT-10 horizontal array survey 
data in a typical 64-hectare field. The following assumptions have been made in the analysis 
shown in Table 17. First, the horizontal array survey is performed on a 12 by 12 grid (yielding 
144 survey sites), two array readings are acquired each site (a 1 metre and 2 metre span), and the 
data acquisition at each site takes 1 minute. Second, the total walking distance and walking time 
is again assumed to be the same as the distance and time required in the conventional sampling 
approach. Third, one hour is needed to perform all the post-survey SCT-10 data processing, and 
both the survey and data processing can be performed by one technical specialist (at a rate 



110 Operational and equipment costs 



US$20/hour), and forth, the total equipment cost is US$2 300, the expected equipment lifespan is 
4 000 hours, and hence the average equipment cost can be estimated to be US$0.5 8/hour. 

The total labour costs associated with this survey would be US$124 and the total 
equipment costs would be US$3.02. Hence, the overall cost associated with the SCT-10 
horizontal array survey would be US$127.02 per field (64-hectare) or US$1.98/ha. 

The EM38 and SCT-10 equipment life spans have been assumed to be 4 000 hours (Tables 
15, 16 and 17). These lifespan estimates are conservative, since they imply that either instrument 
would only last about 100 weeks (i.e., two years) when used 8 hours per day, five days a week. In 
reality, both instruments will last much longer than this, provided they are not seriously abused 
during the field survey work. 

Note that the above overall cost estimates only reflect the costs associated with the 
instrument survey process; the costs associated with the collection of additional calibration soil 
samples still must be determined for two of the survey techniques. For calibration purposes, 
assume that the regression models associated with either the EM38 or SCT-10 horizontal array 
data will be estimated using 12 soil samples. Furthermore, note that one must estimate a unique 
regression model for each sample depth (i.e., 3 separate regression models for the three sample 
depths). Thus, soil samples from 12 separate sample sites must be acquired to calibrate either the 
EM38 or horizontal array data, implying that 36 total soil samples must be analysed. 

Table 18 documents the costs associated with acquiring either EM38 or SCT-10 horizontal 
array calibration soil samples in the 64-hectare field. The assumptions made in Table 18 are 
identical to those made in Table 13, except for the following adjustments: (1) only 12 sites need 
to be sampled, and (2) the post laboratory data processing time can be done in one hour. Hence, 
the revised total labour costs for this calibration sampling become US$112.80, and the revised 
total sampling costs become either US$720 (if the samples are sent to an commercial laboratory) 
or US$83. 16 (if the salinity is measured using the saturated soil paste methodology). 

TABLE 18 

Operational costs for the calibration soil sampling associated with either the EM38 or SCT-10 

horizontal array survey 



Assumptions 800 m by 800 m field size (64 ha); (12 sample sites) 

3 sample depths per site (0-30, 30-60, 60-90 cm depths) 
Time Sampling (15 minutes/site); Walking speed: 4 km/hour. 

Sampling costs External Laboratory US$20/sample 

Internal Laboratory US$2.31 / sample 
Total walking (number of transects +2) x (transect length) = 1 4 x (0.8 km) = 1 1 .2 km 

distance 

Total walking time 11.2/4 = 2.8 hours 

Total sampling time (15 min ) x (12 cores)/60 min = 3 hours 
Data processing time 1 hour 

Labour costs Specialist at US$20/hour and Technical support at US$8/hour 

Crew size 2 technical support for sampling and 1 specialist for data processing 

Total labour costs 5.8 hours @ US$1 6/hour + 1 .0 hours @ US$20/hour = US$1 1 2.80 

Total sampling costs (12 sites) x (3 samples / site) @ US$20/sample = US$720 (using commercial 

laboratory) or 

(12 sites) x (3 samples / site) @ US$2.31/sample = US$83.16 (using 

saturated paste method, internal laboratory) 
Overall survey cost US$832.80 per field (64-hectare) or US$13.01/ha (commercial lab) 
US$195.96 per field (64-hectare) or US$3.06/ha ( internal lab ) 



Soil salinity assessment 



111 



Given these figures, it is now possible to estimate the total (composite) survey costs 
associated with each instrument. Table 19 lists these estimates, which represent the sum total of 
three separate costs: the instrument survey costs, the calibration sampling costs, and the final data 
processing costs. (In Table 19, the data processing time is estimated to be 6 hours regardless of 
which instrumentation approach is employed. This time includes performing all the stochastic or 
deterministic salinity calibrations and generating the final field salinity estimates and/or maps.) 
Note that the overall, total costs for the EM38, SCT-10 horizontal array (internal Lab), and 
SCT-10 insertion four-electrode surveys come to US$413.28, US$442.98, and US$420.57 per 
field (64-hectare), respectively (assuming internal laboratory analysis of all calibration soil 
samples in the first two surveys, and no collection of calibration samples in the third survey). 

This last cost analysis highlights two points. First, on a total expense basis, there is 
relatively little difference between the three types of instrumentation survey techniques. This is 
because the various cost factors (i.e., survey costs, hourly instrument costs, and calibration 
sampling costs) tend to balance each other out. More importantly, these final figures show the 
tremendous cost savings which can be achieved through the judicious use of field instrumental 
methods in conjunction with internal laboratory (saturated soil paste appraisal) methods. Recall 
that the original cost for performing a conventional soil survey within this hypothetical 64-hectare 
field was estimated to be US$9 340.80. By employing the EM38 or either one of the SCT-10 
surveys described above, one could now reduce this cost to about US$415 - US$440. In other 
words, one can achieve a 96% reduction in the overall survey cost. 



TABLE 19 

Total survey and calibration costs for the different survey instrumentation approaches 



Base : Data processing time is 6 hours @ US$20/ hour for either type of survey (includes basic 
statistical or deterministic calibration and map generation) 


Item 

1 . SCT-1 Survey Cost, insertion four-electrode (IF), (Table 16) 

2. SCT-10 Survey Cost, horizontal array (HA), (Table 17) 

3. EM38 Survey Cost, (Table 15) 

4. Calibration Sampling Costs EM38 or SCT-1 0(HA)/Commercial Lab, (Table 18) 

5. Calibration Sampling Costs EM38 or SCT-1 0(HA)/Internal Lab, (Table 18) 

6. Calibration Sampling Costs SCT-1 0(IF)/ no soil samples 

7. Data processing cost 


US$ 
300.57 
127.02 
97.32 
832.80 
195.96 
0.00 
120.00 


Total survey costs = instrument survey cost + calibration sampling 
cost + data processing cost 


US$ per field 
(64-hectare) 


US$/ha 


EM38 / Commercial Lab (Per field = Line 3 + Line 7 + US$ 832.80) 

EM38 / Internal Lab (Per field = Line 3 + Line 7 + US$ 195.96) 

SCT-1 0(HA) / Commercial Lab (Per field = Line 2 + Line 7 + US$ 

832.80) 

SCT-1 0(HA)/ Internal Lab (Per field = Line 2 + Line 7 + US$ 195.96) 

SCT-1 0(IF) / no soil samples (Per field = Line 1 + Line 7) 


1 050.12 

413.28 

1 079.82 

442.98 
420.57 


16.41 

6.46 

16.87 

6.92 
6.57 



Conventional Sampling versus Survey Instrumentation Costs in Multi-Field (Large Area) 
Survey Applications 



The previous cost analysis discussion has been based on a comparison of costs for surveying a 
single field. In practice, a large scale salinity survey will typically encompass many fields (and/or 
a fairly large survey area). Therefore, a comparison of the conventional sampling versus survey 
instrumentation costs inherent in such a large scale survey is obviously desirable. 



112 



Operational and equipment costs 



Table 20 documents the conventional sampling and survey instrumentation costs associated 
with a large scale salinity survey encompassing 6 400 hectares of continuous agricultural land. 
The following assumptions have been made in the analysis shown in Table 20. First, the 6 400- 
hectare area is assumed to be comprised of 100 individual 64-hectare fields, and each field is to 
be sampled (or surveyed) on a 12 by 12 grid. Second, for the conventional sampling, all the time 
and cost assumptions stated in Table 13 are assumed to be the same. Third, for the survey 
instrumentation process, each field is surveyed using an EM38 meter. Hence, all the time and 
instrument survey cost assumptions stated in Table 15 are assumed to be the same. Additionally, 
it is assumed that 4 sites from each field are selected for calibration soil sampling (generating 400 
sample sites and hence 1 200 soil samples across the survey area), and that this calibration 
sampling cost can be estimated to cost US$72.92 on a per field basis (see Table 21). Third, an 
additional cost of US$3 000 (for purchasing a single GPS unit) is assumed to be incurred by 
either survey process. Fourth, for the survey instrumentation process, an additional cost of 
US$4 000 is incurred for 80 hours worth of geostatistical analysis (for purposes of conductivity- 
to-salinity calibration and map generation). And finally, the conventional sampling approach uses 
a commercial laboratory to perform all of the laboratory salinity analyses, and the calibration soil 
samples are analysed for salinity using the saturation paste conductivity method (i.e., internal 
laboratory). 

TABLE 20 

Cost comparison between the conventional versus survey instrumentation and calibration 

methodologies for a 6 400-hectare survey (one hundred 64-ha fields). 



Assumptions: (i)1 00 fields, each 800 m by 800 m, all located across one continuous area, (ii) 12 by 12 
grid (144 sites) per field, (iii) Conventional approach requires 144 sample sites (3 samples per depth), 
(iv) Composite approach requires 144 survey sites (2 readings per site) + 4 sample sites for calibration 
(3 sample per depth), (v) Both approaches require the purchase of 1 GPS unit @ US$3,000 per unit, 
(vi) Conventional approach uses commercial laboratory and (vii) Composite approach uses internal 
laboratory (saturated paste method) 



Total conventional survey cost 



Total instrument survey cost 

Total calibration sampling cost 

GPS cost 

Total data analysis cost (for geostatistical analysis) 

Total instrument Survey/Calibration/Analysis Cost 



= 100 x (US$9 340.80) + GPS cost 

= US$934 080 + 3,000 = US$937 

080 

100 x (US$97.32) = US$9,732 

100 x (US$72.92) = US$7,292 

US$3,000 

80 hours @ US$50/hour = US$4 000 

US$24 024 



Notes: (i) Overall time to complete conventional survey is 4,280 hours, not including external 
laboratory time for salinity analysis and (ii) Overall time to complete instrument survey / calibration 
analysis is 950 hours, including internal laboratory salinity analysis and contracted data analysis. 

The overall cost of the conventional sampling approach would be US$937 080 
(US$146.42/ha), while the overall cost of the survey instrumentation approach would be only 
US$24 024 (US$3.75/ha). Hence, by exploiting the use of the survey instrumentation approach, 
the composite cost of the entire survey process would be reduced by over 97%. 



Some details inherent to the assumptions in Table 20 are worth expanding on. It has been 
assumed that EM38 instrumentation is being used in this analysis, because this survey process is 
the fastest (and hence the least expensive). However, for calibration purposes, only four samples 
per field are acquired. The reason why this further reduction in the number of soil samples is 
possible is due to the fact that the 6 400-hectare survey area is assumed to be continuous. Hence, 
because the survey has been conducted across one large area, it will be possible (and in fact, 
preferable) to employ more sophisticated geostatistical techniques to estimate the calibration 
equation(s). Note that such an analysis would have to typically be performed by a trained 



Soil salinity assessment 113 



statistician, implying an additional data analysis expense (in this example, US$4 000). However, 
this additional expense is more than offset by the cost savings achieved through the reduction of 
the soil sampling requirements. 

Second, the total survey time under the survey instrumentation approach is significantly 
faster than the total time under the conventional sampling approach. The instrumentation 
approach requires a total of 950 hours; which represents the sum of 340 EM38 survey hours, 220 
soil sampling hours, 150 data processing hours, 160 laboratory analysis hours, and 80 hour of 
external data analysis. In contrast, the conventional sampling approach would require a total of 
4 280 hours, not including the time required for the commercial laboratory analysis of the soil 
samples. Additionally, since 340 hours < 4 000 (equipment life span), all of the EM38 survey 
work can be done with one EM38 and since 1 200 (number of calibration samples) < 15 000 
(equipment life span in sample number) all of the saturated paste soil conductivity measurements 
can be made with one CO 150 conductivity meter. 

Finally, it should be pointed out that this sort of non-invasive survey / calibration sampling 
process is not necessarily the most cost effective approach if the survey region is highly 
discontinuous and/or the survey grid is extremely large. If 100 separate, discontinuous fields are 
to be surveyed, then in general no reduction in the per field calibration sampling size is possible. 
Hence, the cost differences between the three survey instrumentation techniques are rather 
minimal (see Table 19). Likewise, suppose a very coarse survey grid was to be employed over a 
large area; for example, a 1 km by 1 km grid over a 10 000-km 2 area. Then the variation in the 
secondary soil physical properties would undoubtedly be quite large and the site to site 
correlation between these properties would probably be minimal. Hence, the insertion four- 
electrode technique would be expected to yield the most accurate information for the least cost, 
since the calibration sampling requirements for the other two techniques would be greatly 
increased. 

TABLE 21 

Operational costs for the calibration soil sampling associated with 6 400-hectare survey 

discussed in Table 20 



Assumptions 800 m by 800 m field size (64 ha); (4 sample sites per field) 

3 sample depths per site (0-30, 30-60, 60-90 cm depths) and internal 
laboratory (saturated paste) appraisal methods. 
Time Sampling (15 minutes/site); Walking speed: 4 km/hour. 

Sampling costs Internal Laboratory US$2.31 / sample 

Total walking distance (number of transects +2) x (transect length) 6 x (0.8 km) = 4.8 km 

Total walking time 4.8 / 4.0 = 1 .2 hours 

Total sampling time (15 minutes/site) x (4 sites) / 60 min = 1 hour 

Data processing time 0.5 hour 

Labour costs Specialist at US$20/hour and Technical support at US$8/hour 

Crew size 2 Technical support for sampling, 1 Specialist for data processing 

Total labour costs 2.2 hours @ US$1 6/hour + 0.5 hours @ US$20/hour = US$45.20 

Total sampling costs (4 sites) x (3 samples / site) @ US$2.31 / sample US$27.72 (using 

saturated paste method, internal laboratory) 
Overall survey cost US$72.92 per field (64-hectare) or US$1.14/ha 



Cost Advantages associated with Instrument Mobilization 

As the previous example demonstrates, it is possible to survey fairly large areas without 
mobilizing any of the instrumentation equipment. However, as the survey regions become 
increasingly large, the additional cost saving advantages of instrument mobilization can become 
quite significant. 



114 



Operational and equipment costs 



There are three primary cost advantages associated with the mobilization of the various 
salinity survey instruments. The first advantage is the speed in which the survey process can be 
completed. Mechanized systems can almost always be used to survey more land than hand held 
systems, simply because of their increased travel speed. Second, mechanized systems can be used 
to collect significantly more survey data. Indeed, many of the commercial systems currently 
available can collect survey data in a nearly continuous fashion. And third, when hand held units 
are mounted or adapted into mobilized platforms, they simply tend to last longer. This increased 
lifespan is primarily due to the significant reduction in operator handling and/or manual abuse 
that these units receive. 

TABLE 22 

Typical start up costs for three types of mobilized instrument assessment systems 



Dual EM38 Trailering System 


US$ 


EM38 Meters (2x US$7,195) 

DL720 data logger 

Trailer Platform 

Instrument interfacing 

GPS Unit 

Alternative terrain vehicle (ATV) 

Total system cost 


14 390 
3 775 

3 000 

4 500 
3 000 
6 000 

34 665 


Expected system Life 
Equipment maintenance 

Cost per hour 


10,000 hours 

50%(ATV cost) + 10%(Trailer cost) 

US$3 300 over expected system life 

(34 665 + 3 300) / 10 000 = US$3.80/hour 


Mobilized, Tractor Mounted Horizontal Array 
System 

Horizontal array system 

GPS Unit 

Tractor 

Total system cost 


US$ 

5 250 

3 000 

12 500 

20 750 


Expected System Life 
Equipment Maintenance 

Cost per hour 


10 000 hours 

50%(Tractor cost) 

US$6 250 over expected system life 

(20 750 + 6 250)/10 000 = US$2.70 per hour 


Mobilized, Multi-Instrument Transport System 

Basic hydraulic transport vehicle 

Structural fabrication costs 

EM38 Meter 

DL720 data logger 

SCT-10 Meter 

GPS Unit 

Instrument interfacing 

Total system cost 

Expected system life 

Equipment Maintenance 

Cost per hour 


us$ 

8 000 

27 933 

7 195 

3 775 

2 000 

3 000 
1 400 

53 303 
10 000 hours 

50%(Basic Transport Vehicle cost) + 10% 
(Fabrication cost) = US$6 793 over expected 
system life 
(53,303 + 6,793) / 10,000 = US$6.01 per hour 



Notes: secondary vehicle costs (for transporting systems from field to field) are not included in these 
cost per hour estimates. 



Table 22 lists the approximate start up costs for three mechanized salinity assessment 
systems; a dual EM38 trailering system, a mobilized horizontal array system, and a mobilized, 
multi-instrument transport system. The first two systems can be used to rapidly acquire bulk soil 
conductivity information in a continuous manner, while the third system is primarily designed to 
rapidly collect more depth specific conductivity information in a stop and go manner. Hence, the 
dual EM38 trailering system and mobilized horizontal array system are ideally suited for 



Soil salinity assessment 115 



collecting large amounts of average soil profile conductivity information. In contrast, the 
mobilized, multi- instrument transport system is ideally suited for rapidly acquiring multiple sets 
of soil conductivity data over a grid of fixed locations. Hence, this system supplies more 
accurate, depth specific information, which in turn makes it more useful for repetitive salinity 
monitoring activities and/or any large scale inventorying process with requires high vertical (soil 
profile) resolution. 

The total start up cost for the dual EM38 trailering system includes the following: two 
EM38 units (one each for horizontal and vertical readings), one DL720 data logger, the trailer 
platform (estimated to cost US$3 000), the electronic instrument interfacing (estimated to cost 
US$4 500), one GPS unit, and one ATV. Assuming that the GPS unit and ATV can be 
purchased for US$3 000 and US$6 000, respectively, the total system cost would be US$34 665. 
Additional equipment maintenance expenses (for maintaining the ATV and trailer) should also be 
factored into the system cost; in Table 22 this cost has been estimated to be approximately 
US$3 300 over the life of the system. Assuming that the system life is 10 000 hours (which is 
roughly equivalent to 5 years, if the system is operated 8 hours a day, 5 days a week), the average 
cost per hour for operating this dual EM38 trailering system would be US$3.80. Note that this 
hourly cost does not include any fuel or labour costs associated with the surveying process. 

The total start up cost for the tractor mounted, horizontal array system would be 
US$20 750, after factoring in the cost of the horizontal array unit, the GPS unit, and a tractor. 
The additional equipment maintenance cost for maintaining a tractor would probably be 
somewhat higher; in Table 22 it has been estimated to be approximately US$6 250 over the life 
of this system. Thus, assuming the same sort of system life expectancy (10 000 hours), the 
average cost per hour for operating this system would be US$2.70. Once again, this hourly cost 
does not include any fuel or labour costs associated the surveying process. 

In a similar manner, the total start up cost for the multi-instrument transport system would 
be US$53 303, after factoring in all of the various equipment costs, structural fabrication costs, 
and the base cost of the initial hydraulic transport vehicle. Additionally, the estimated equipment 
maintenance costs for this system would be US$6 793 of the expected system life. Again, 
assuming a system life of 10 000 hours, the average cost per hour for this system would be 
US$6.01 (not including fuel or labour costs). 

Note that in all three start up cost examples, no additional secondary vehicle costs have 
been incorporated into the analysis. However, some sort of towing vehicle would typically be 
needed if any of these systems were to be transported over significant distances. Additionally, the 
start up costs for both the dual EM38 and the tractor mounted horizontal array systems could be 
significantly reduced, if one was already in possession of a suitable ATV or tractor. 

Based on the hourly cost estimates discussed above, it is now possible to calculate the total 
instrument survey cost one would incur in a survey of a typically 64-hectare field using each of 
these three systems. These survey costs are documented in Tables 23, 24 and 25 for the dual 
EM38, horizontal array, and multi-instrument transport systems, respectively. 

The assumptions used in the Table 23 calculations are as follows. The dual EM38 
trailering system is assumed to traverse the field at a rate of 12 km per hour, and the survey is 
performed across 24 equally spaced 800 meter transects. Thus, the total travel distance would be 
20.8 km, and the system fuel costs are assumed to be US$0.15 per km. Furthermore, the system 



116 



Operational and equipment costs 



can still be operated by one technical specialist (paid at a rate of US$20/hour), and the post 
survey data processing still takes one hour. Based on these assumptions, the overall mechanized 
EM38 survey cost for this 64-hectare field would come to US$64.29, or about US$1.00 per 
hectare. 



TABLE 23 

Operational survey costs associated with a mobilized EM38 survey in a typical 64-hectare field 

(dual EM38 trailering system costs) 



Assumptions 


800 m by 800 m field size (64 ha) 




24 transects (readings collected every 1 to 10 seconds per 




transect ) 




(576 to 5760 total survey sites) 


Cost per hour 


US$3.80 


Travel time 


12 km / hour 


Total travel distance 


(number of transects +2) x (transect length) 26 x (0.8 km) = 20.8 

km 

20.8 / 1 2.0 = 1 .73 hours (includes survey time) 


Total travel time 


Total fuel costs 


(US$0.15 per km) x 20.8 km = US$3.12 


Data processing time 


1 hour 


Labour costs 


Specialist at US$20/hour 


Crew size 


1 Specialist for automated survey work and data processing 


Total labour costs 


2.73 hours @ US$20/hour = US$54.60 


Total equipment costs 


1 .73 hours @ US$3.80 per hour + fuel cost = US$9.69 


Overall survey cost 


US$64.29 per field (64-hectare) or US$1/ha. 



A nearly identical set of assumptions in Table 24 yield an overall horizontal array survey 
cost of US$83.18, or about US$1.30 per hectare. Note that the only differences in the Table 24 
assumptions concern the travel time and fuel costs. Direct contact horizontal array systems tend 
to be towed across a field at a somewhat slower rate, and a tractor would be expected to use 
slightly more fuel. 



TABLE 24 

Operational survey costs associated with a mobilized, horizontal array conductivity survey in a 

typical 64-hectare field 



Assumptions 


800 m by 800 m field size (64 ha); 24 transects (readings collected 




every 1 to 10 seconds per transect) and 864 to 8640 total survey sites 


Cost per hour 


US$2.70 


Travel time 


8 km/hour 


Total travel distance 


(number of transects +2) x (transect length) = 26 x (0.8 km) = 20.8 km 


Total travel time 


20.8 / 8.0 = 2.6 hours (includes survey time) 


Total fuel costs 


(US$0.20 per km) x 20.8 km = US$4.16 


Data processing time 


1 hour 


Labour costs 


Specialist at US$20/hour 


Crew size 


1 Specialist for automated survey work and data processing 


Total labour costs 


3.6 hours @ US$20/hour = US$72.00 


Total equipment costs 


2.6 hours® US$2.70 per hour + fuel cost = US$11.18 


Overall survey cost 


US$83.18 per field (64-hectare) or US$1.30/ha 



The assumptions shown in Table 25 are slightly different. A mobilized, multi-instrument 
transport system would generally be used to perform a grid survey similar to the earlier manual 
survey instrumentation processes (only faster and in more detail). Hence, in Table 25 the same 12 
by 12 survey grid is used. Additionally, 4 EM38 and 4 horizontal array survey readings are 
acquired at each site, and the total data acquisition time is assumed to take 45 seconds (per site). 
As in Table 23, the travel time is assumed to be 12 km/hour, the fuel costs are assumed to be 
US$0.15/km, one technical specialist can still operate the system, and data processing still takes 



Soil salinity assessment 117 



one hour. Based on these assumptions, the overall multi-instrument transport system survey cost 
for this 64-hectare field would come to US$92.69, or about US$1.45/ha. 

TABLE 25 

Operational survey costs associated with a mobilized, multi-instrument survey in a typical 64- 

hectare field (multi-instrument transport system costs) 



Assumptions 


800 m by 800 m field size (64 ha); 12 by 12 grid (144 survey 
4 EM38 and horizontal array survey readings per site. 


sites) and 


Cost per hour 


US$6.01/hour 




Travel time 


12 km/hour 




Survey time 


45 seconds per site 




Total travel distance 


(number of transects +2) x (transect length) = 14 x (0.8 km) = 


= 11.2 km 


Total travel time 


11.2/12.0 = 0.93 hours 




Total survey time 


(0.75 min) x (144 sites)/60 min = 1.8 hours 




Total fuel costs 


(US$0.15 per km) x 11.2 km = US$1.68 




Data processing time 


1 hour 




Labour costs 


Specialist at US$20/hour 




Crew size 


1 Specialist for automated survey work and data processing 




Total labour costs 


3.73 hours @ US$20/hour = US$74.60 




Total equipment costs 


2.73 hours @ US$6.01 per hour + fuel cost = US$18.09 




Overall survey cost 


US$92.69 per field (64-hectare) or US$1.45/ha 





In all three cases, the final per hectare costs can be seen to be lower than the lowest manual 
instrument survey cost, which was US$1.52 per hectare for the EM38 (Table 15). Furthermore, 
the amount of instrument survey data collected during each of the automated survey processes 
has been greatly increased. For example, given the assumptions in Table 23, the dual EM38 
trailering system would complete each survey transect in 4 minutes. If one set of readings 
(horizontal and vertical) are acquired every 5 seconds, then this automated system can effectively 
collect signal data at 1152 survey sites within the field (24x240/5=1 152). Hence, either the dual 
EM38 or tractor mounted horizontal array systems can be used to greatly increase the spatial 
resolution of the conductivity data without raising the per hectare survey instrumentation costs. 
Likewise, the multi-instrument transport system can effectively collect 4 to 8 times the amount of 
conductivity data in the same amount of time over each survey site, compared to manual survey 
methods. Thus, the vertical resolution of the survey process is greatly increased, again without 
raising the per hectare survey instrumentation costs. 

For very large survey applications, instrument mobilization becomes highly cost effective. 
For example, each of these mobilized systems could be used to effectively survey thousands of 
hectares per week in typical, large scale agricultural regions. Without question, this represents a 
far greater amount of landscape than can ever be effectively inventoried in the same amount of 
time using manual survey techniques. 



Conclusion 

A comprehensive cost analysis of the various survey instrumentation techniques currently used 
for soil salinity assessment has been carried out. It included a description of the capital costs 
associated with both manual and mobilized instrumentation techniques, useful analytical 
laboratory techniques, and current software. Additionally, a detailed analysis of the operational 
costs incurred when applying these various techniques for the measurement of soil salinity has 
been performed. 



118 Operational and equipment costs 



Table 26 summarizes the operational costs associated with the various survey 
instrumentation techniques discussed herewith. The operational cost of the three most common 
manual instrumentation techniques were all found to be between US$6.40 to US$6.90 per hectare 
in a typical 64-ha field. These operational cost figures were comprised of three components; the 
actual survey instrumentation costs, the data analysis costs, and the calibration sampling costs 
(when necessary). In all three cases, these figures were about 96% less than the conventional soil 
sampling costs, which were estimated to be about US$146/ha. Additionally, the cost analysis of 
the three most common mechanized instrumentation systems demonstrated that these per hectare 
costs can be further reduced by suitably mobilizing the various instruments, especially when large 
agricultural regions are to be surveyed. 

Overall, this cost benefit analysis suggests that the proper implementation of the various 
survey instrumentation techniques can lead to significant financial savings in most all types of 
salinity survey applications. Furthermore, these savings would be expected to be greatest in large 
scale agricultural surveys and/or regional inventorying applications. 

TABLE 26 

Summary table of various survey instrumentation costs discussed in Tables 13 through 25. All 
cost and time factors based on a 64-hectare field size, and time factors do not include 
commercial or internal laboratory analysis time components 



Survey Method 




Cost, 






Time 






Number of Survey 






US$/ha 






hours/field 




Sites per field 


Individual 64-ha Field 


















Conventional sampling (1> 




145.95 






42.8 






144 


Conventional sampling (2) 




26.54 






42.8 






144 


EM38 manual ,3) 




6.46 






4.4 






144 


SCT-10 HA manual ,3) 




6.92 






6.2 






144 


SCT-10 IF manual ,3) 




6.57 






14.6 






144 


EM38 mobilized (4) 




5.94 






2.7 






576+ 


SCT-10 HA mobilized (4) 




6.24 






3.6 






576+ 


Multi-instrument mobilized (4 




6.39 






3.7 






144 


Continuous 6400 ha Region: 
















EM38 mobilized (5) 




3.24 






2.7 






576+ 


SCT-10 HA mobilized' 51 




3.53 






3.6 






576+ 


Multi-instrument mobilized (5 




3.68 






3.7 






144 


Notes: (1) from Table 13; 


(2) from 


Table 14; 


[3) 


from 


Table 19; 


(4) 


from 


Tables 23-25 & 19 


(substituting mobilized for 


manual si 


rvey costs] 


and (5) 


from Tables 


23-25 


& 20 (substituting 


mobilized for manual survey 


costs). 

















Soil salinity assessment 119 



References 



Adams, F. 1974. Soil solution. In E.W. Carson (ed.) The plant root and its environment. Univ. Press of 
Virginia, Charlottesville, p. 441-481. 

Adams, F., C. Burmester, N. V. Hue, and F. L. Long. 1980. A comparison of column -displacement and 
centrifuge methods for obtaining soil solutions. Soil Sci. Soc.Am. J. 44:733-735. 

Adiku, S. G. K., M. Renger, and C. Roth. 1992. A simple model for extrapolating the electrical 
conductivity data of gypsum containing soils from reference soil extract data. Agri. Water Mgt. 
21:235-246. 

Allison, L.E. 1973. Oversaturation method for preparing saturation extracts for salinity appraisal. Soil 
Sci. 116:65-69. 

Austin, R. S., and J. D. Oster. 1973. An oscillator circuit for automated salinity sensor measurements. 
Soil Sci. Soc. Am. Proc. 37:327-329. 

Austin, R.S., and J. D. Rhoades. 1979. A compact, low-cost circuit for reading four-electrode salinity 
sensors. Soil Sci. Soc. Am. J. 43:808-810. 

Barnes, H. E. 1954. Electrical subsurface exploration simplified. Roads and Streets, May 1954, pp. 81- 
84,140. 

Beatty, H.J., and J. Loveday. 1974. Soluble cations and anions. In J. Loveday (ed.) Methods for analysis 
of irrigated soils. Technical Communication No. 54, Commonwealth Bureau of Soils, 
Commonwealth Agricultural Bureau, Farmham Royal, Bucks, England, p. 108-117. 

Biggar, J. W., and D. R. Nielsen. 1976. Spatial variability of the leaching characteristics of a field soil. 
Water Resour. Res. 12:78-84. 

Bohn, H.L., J. Ben-Asher, H.S. Tabbara, and M. Marwan. 1982. Theories and tests of electrical 
conductivity in soils. Soil Sci. Soc. Am. J. 46:1143 — 1146. 

Bottraud, Jean-Christophe, and J. D. Rhoades. 1985a. Effect of exchangeable sodium on soil electrical 
conductivity-salinity calibrations. Soil Sci. Soc. Am. J. 49:1110-1113. 

Bottraud, Jean-Christophe, and J. D. Rhoades. 1985b. Referencing water content effects on soil electrical 
conductivity - salinity calibrations. Soil Sci. Soc. Am. J.49: 1579-1581. 

Briggs, L.J., and A.G. McCall. 1904. An artificial root for inducing capillary movement of soil moisture. 
Sci. 20:566-569. 

Cameron, D. R., E. De Jong, D.W.L. Read, and M. Oosterveld. 1981. Mapping salinity using resistivity 
and electromagnetic inductive techniques. Can. J. Soil Sci. 61:67-78. Cameron, M. E., R. C. 
McKenzie and G. Lachapelle, 1994. Soil salinity mapping with electromagnetic induction and 
satellite-based navigation methods. Can. J. Soil Sci. 74:335-343. 

Carter, L. M., J. D. Rhoades and J. H. Chesson. 1993. Mechanization of soil salinity assessment for 
mapping. Proc, 1993 ASAE Winter Meetings, Chicago, IL., December 12-17, 1993 

Chow, T. L. 1977. A porous cup soil-water sampler with volume control. Soil Sci. 124:173-176. 



1 20 References 



Cook, P. G., M. W. Hughes, G. R. Walker and G. B. Allison. 1989. The calibration of frequency-domain 
electromagnetic induction meters and their possible use in recharge studies. J. Hydrology 107:251- 
265. 

Corwin, D.L., and J. D. Pdioades. 1982. An improved technique for determining soil electrical 
conductivity-depth relations from above ground electromagneticm measurements. Soil Sci. Soc. 
Am. J. 46:517-520. 

Corwin, D.L., and J. D. Pdioades. 1984. Measurement of inverted electrical conductivity profiles using 
electromagnetic induction. Soil Sci. Soc. Am. J. 48:288-291. 

Corwin, D.L., and J. D. Rhoades. 1990. Establishing soil electrical conductivity - depth relations from 
electromagnetic induction measurements. Commun. Soil Sci. Plant Anal. 21:861-901. 

Dao, T.H., and T. L. Lavy. 1978. Extraction of soil solution using a simple centrifugation method for 
pesticide adsorption - desorption studies. Soil Sci. Soc. Am. J. 42:375-377. 

Davies, B.E., and R. I. Davies. 1963. A simple centrifugation method for obtaining small samples of soil 
solution. Nature. 198:216-217. 

Diaz, L. and J. Herrero. 1992. Salinity estimates in irrigated soils using electromagnetic induction. Soil 
Sci. 154:151-157. 

Dobrin, M. B. 1960. Introduction to Geophysical Prospecting, McGraw-Hill Book Company, 446 pp. 

Doolittle, J. A., K.A. Sudduth, N. R. Kitchen, and S. J. Indorante. 1994. Estimating depths to claypans 
using electromagnetic induction methods. J. Soil and Water Cons. 49:572-575. 

Duke, H.R., and H.R. Haise. 1973. Vacuum extractors to assess deep percolation losses and chemical 
constituents of soil water. Soil Sci. Soc. Amer. Proc. 37:963-964. 

Elkhatib, E.A., O.L. Bennett, V.C. Baligar, and R.J. Wright. 1986. A centrifuge method for obtaining 
soil solution using an immiscible liquid. Soil Sci. Soc. Am. J. 50:297-299. 

Elkhatib, E.A., J.L. Hern, and T.E. Staley. 1987. A rapid centrifugation method for obtaining soil 
solution. Soil Sci. Soc. Am. J. 51:578-583. 

Enfield, C.G., and D. D. Evans. 1969. Conductivity instrumentation for in situ measurements of soil 
salinity. Soil Sci. Soc. Amer. Proc. 33:787-789. 

FAO. 1988. World Agriculture Toward 2000: An FAO Study. N. Alexandratos (ed.). Bellhaven Press, 
London. 338 pp. 

Gerlach, F. L., and A. E. Jasumbach. 1989. Global positioning systems canopy effect study. Technol. 
Development Center, Forest Service, U.S. Dept. Agr., Missoula, MT., 18 pp. 

Ghassemi, F., A. J. Jakeman and H. A. Nix. 1995. Salinization of Land and Water Resources: Human 
Causes, Extent, Management and Case Studies, CAB International, 526 pp. 

Gillman, G. P. 1976. A centrifuge method for obtaining soil solution. Div. Rep. No. 16. CSIRO, Div. of 
Soils, Townsville, Queensland, Australia. 

Grieve, A.M., E. Dunford, D. Marston, RE. Martin, and P. Slavich. Effects of waterlogging and soil 
salinity on irrigated agriculture in the Murray Valley: a review. Australian Journal of 
Experimental Agriculture . 1986, 26, 761-77 '. 

Grossman, J., and P. Udluft. 1991. The extraction of soil water by the suction-cup method: A review. J. 
Soil Sci. 42:83-93. 

Halvorson, A. D., and J. D. Rhoades. 1974. Assessing soil salinity and identifying potential saline-seep 
areas with field soil resistance measurements. Soil Sci. Soc. Amer. Proc. 38:576-581. 

Halvorson, A. D., and J. D. Rhoades. 1976. Field mapping soil conductivity to delineate dryland saline 
seeps with four-electrode technique. Soil Sci. Soc. Am. J. 40:571-575. 



Soil salinity assessment 121 



Halvorson, A. D., J. D. Rhoades, and C.A. Ruele. 1977. Soil salinity-four-electrode conductivity 
relationships for soils of the Northern Great Plains. Soil Sci. Soc.Am. J. 41:966-971. 

Harris, Alfred Ray, and Edward A. Hansen. 1975. A new ceramic cup soil-water sampler. Soil Sci. Soc. 
Amer. Proc. 39:157-158. 

Heimovaara, T. J. 1995. Assessing temporal variations in soil water composition with time domain 
reflectometry. Soil Sci. Soc. Am. J. 59:689-698. 

Hoosbeek, M. R., A. Stein, and R. B. Bryant. 1997. Mapping soil degradation. In R. Lai, W. H. Blum, 
C. Valentine, and B. A. Stewart (Eds.), Methods for Assessment of Soil Degradation, CRC Press, 
New York, pp. 407-422. 

Ingvalson, R. D., J. D. Oster, S.L. Rawlins, and G.J. Hoffman. 1970. Measurement of water potential 
and osmotic potential in soil with a combined thermocouple psychrometer and salinity sensor. Soil 
Sci. Soc. Amer. Proc. 34:570-74. 

Jackson, D. R., F.S. Brinkley, and E.A. Bendetti. 1976. Extraction of soil water using cellulose-acetate 
hollow fibers. Soil Sci. Soc. Am. J. 40:327-329. 

Jaynes, D.B. 1996. Improved soil mapping using electromagnetic induction surveys. In P.C. Robert, 
R.H. Rust, and W.E. Larson (ed.) Proc. 3rd Inter. Conf. on Precision Agriculture. Minneapolis, 
MN. June 23-26, 1996. ASA, CSSA, SSSA, Madison, WI. 

Johnston, M. A. 1994. An evaluation of the four-electrode and electromagnetic induction techniques of 
soil salinity measurement. Water Research Commission Report No. 269/1/94, South Africa, pp. 
191. 

Jordan, Carl F. 1968. A simple, tension-free lysimeter. Soil Sci. 105:81-86. 

Jury, W. A. 1975a. Solute travel-time estimates for tile-drained fields: I. Theory. Soil Sci. Soc. Am. J. 
39:1020-1024. 

Jury, W. A. 1975b. Solute travel-time estimates for tile-drained fields. II. Application to experimental 
studies. Soil Sci. Soc. Am. J. 39:1024-1028. 

Kaddah, M. T. and J. D. Rhoades. 1976. Salt and water balance in Imperial Valley, California. Soil Sci. 
Soc. Am. J. 40:93-100. 

Kachanoski, R. G., E. G. Gregorich and I. J. van Wesenbeeck. 1988. Estimating spatial variations of soil 
water content using noncontacting electromagnetic inductive methods. Can. J. Soil Sci. 68: 715- 

722. 

Kemper, W. D. 1959. Estimation of osmotic stress in soil water from the electrical resistance of finely 
porous ceramic units. Soil Sci. 87:345-349. 

Kinniburgh, D.G., and D.L. Miles. 1983. Extraction and chemical analysis of interstitial water from 
soils and rocks. Environ. Sci. and Technol. 17:362-368. 

Kitchen, N. R., K. A. Sudduth and S. T. Drummond. 1996. Mapping of sand deposition from 1993 
Midwest floods with electromagnetic induction measurements. J. Soil and Water Cons. 51(4):336- 
340. 

Kittrick, J. A. 1983. Accuracy of several immiscible displacement liquids. Soil Sci. Soc.Am. J. 47:1045- 
1047. 

Kohnke, H, F.R. Dreibelbis, and J.M. Davidson. 1940. A survey and discussion of lysimeters and a 
bibliography on their construction and performance. U.S. Dept. Agr. Misc. Publ. No. 372. 

Lesch, S.M., J. D. Rhoades, L.J. Lund and D.L. Corwin. 1992. Mapping soil salinity using calibrated 
electromagnetic measurements. Soil Sci. Soc. Am. J. 56(2):540-548. 



1 22 References 



Lesch, S.M., D. J. Strauss and J. D. Rhoades. 1995a. Spatial prediction of soil salinity using 
electromagnetic induction techniques: 1. Statistical prediction models: A comparison of multiple 
linear regression and cokriging. Water Resour. Res. 31:373-386. 

Lesch, S. M., D. J. Strauss and J. D. Rhoades. 1995b. Spatial prediction of soil salinity using 
electromagnetic induction techniques: 2. An efficient spatial sampling algorithm suitable for 
multiple linear regression model identification and estimation. Water Resour. Res. 31: 387-398. 

Lesch, S.M., J. D. Rhoades, D. J. Strauss, K. Lin and M.A.A. Co. 1995c. The ESAP user manual and 
tutorial guide version 1.0. U.S. Salinity Laboratory Research Report No. 138, 108 p. 

Lesch, S. M., J. Herrero, and J. D. Rhoades. 1998. Monitoring for temporal changes in soil salinity 
using electromagnetic induction techniques. Soil Sci. Soc. Amer. J. 62:232-242. 

Levin, M.J., and D. R. Jackson. 1977. A comparison of in situ extractors for sampling soil water. Soil 
Sci. Soc. Am. J. 41:535-536. 

Litaor, M. Iggy. 1988. Review of soil solution samplers. Water Resour. Res. 24:727-733. 

Long, D.S., S. D. DeGloria, and J.M. Galbraith. 1991. Use of the global positioning system in soil 
survey. J. Soil Water Conserv. 46:293-297. 

Longenecker, D. E., and P.J. Lyerly. 1964. Making soil pastes for salinity analysis: A reproducible 
capillary procedure. Soil Sci. 97:268-275. 

Lopez-Bruna, D. and J. Herrero. 1996. The behavior of the electromagnetic sensor and its calibration for 
soil salinity. Agronomie 16:95-105. (in Spanish) 

Loveday, J. 1972. Moisture content of soils for making saturation extracts and effect of grinding. 
Division of Soils Technical Paper No. 12a, Commonwealth Scientific and Industrial Research 
Organization, Canberra City, A.C.T. 2601, Australia. 

Loveday, J. 1980. Experiences with the 4-electrode resistivity technique for measuring soil salinity. Div. 
Soils, CSIRO, Rep. 51. 

Luke, R. and P. Shaw. 1994a. An economic assessment of the cost of salinity to agriculture in the 
Campaspe dryland sub-region. Victorian Department of Agriculture, East Melbourne, Australia. 40 
pp. 

Luke, R. and P. Shaw. 1994b. An economic assessment of the cost of salinity to agriculture in the 
Loddon dryland sub-region. Victorian Department of Agriculture, East Melbourne, Australia. 40 
pp. 

Luke, R. and P. Shaw. 1994c. An economic assessment of the cost of salinity to agriculture in the Avoca 
dryland sub-region. Victorian Department of Agriculture, East Melbourne, Australia. 44 pp. 

Maas, E.V. 1986. Salt tolerance of plants. Applied Agricultural Research. 1:12-26. 

Maas, E.V. 1990. Chapter 13. Crop salt tolerance. In K.K. Tanji (ed.) Agri. salinity assessment and 
management, ASCE Manuals & Reports on Engineering No. 71. ASCE, N. Y., pp. 262-304. 

Maas, E.V., and G.J. Hoffman. 1977. Crop salt tolerance - current assessment. J. Irrig. and Drainage 
Div., ASCE 103(IR2): 115-134. 

McKenzie, R.C., W. Chomistek, and N. F. Clark. 1989. Conversion of electromagnetic inductance 
readings to saturated paste extract values in soils for different temperature, texture, and moisture 
conditions. Can. J. Soil Sci. 69:25-32. 

McKenzie, R.C., D. R. Bennett, and K.M. Riddel. 1990. Mapping salinity to measure potential yield of 
wheat. Annual Alberta Soil Science Workshop, Edmonton, Canada, pp. 268-273. 

McNeill, J. D. 1980. Electromagnetic terrain conductivity measurement at low induction numbers. 
Technical Note TN-6, Geonics Limited, Mississauga, Ontario, Canada. 



Soil salinity assessment 1 23 



McNeill, J. D. 1990. Personal communication. 

McNeill, J. D. 1992. Rapid, accurate mapping of soil salinity by electromagnetic ground conductivity 
meters. In G.C. Topp, W. D. Reynolds and R. E. Green (eds.) Advances in measurement of soil 
physical properties: Bringing theory into practice. SSSA Spec. Publ. 30. ASA, CSSA and SSSA, 
Madison, WL, pp. 209-229. 

Menzies, N. W., and L.C. Bell. 1988. Evaluation of the influence of sample preparation and extraction 
technique on soil solution composition. Aust. J. Soil Res. 26:451-464. 

Mubarak, A., and RA. Olsen. 1976. Immiscible displacement of the soil solution by centrifugation. Soil 
Sci. Soc. Am. J. 40:329-331. 

Mubarak, A., and RA. Olsen. 1977. A laboratory technique for appraising in situ salinity of soil. Soil 
Sci. Soc. Am. J. 41:1018-1020. 

Nadler, A. 1981. Field application of the four-electrode technique for determining soil solution 
conductivity. Soil Sci. Soc. Am. J. 45:30-34. 

Nadler, A. 1982. Estimating the soil water dependence of the electrical conductivity soil 
solution/electrical conductivity bulk soil ratio. Soil Sci. Soc. Am. J. 46:722-726. 

Nadler, A., and S. Dasberg. 1980. A comparison of different methods for measuring soil salinity. Soil 
Sci. Soc. Am. J. 44:725-728. 

Nadler, A., M. Magaritz, Y. Lapid, and Y. Levy. 1982. A simple system for repeated soil resistance 
measurements at the same spot. Soil Sci. Soc. Am. J. 46:661-663. 

Nielsen, D. R., J.W. Biggar, and K.T. Fah. 1973. Spatial variability of field-measured soil-water 
properties. Hilgardia. 42:215-259. 

Oldeman, L.R., van Engelen, V.W.P., and Pulles, J.H.M. 1991. The extent of human-induced soil 
degradation. In. Oldeman, L.R., Hakkeling, R.T.A., and Sombroek, W.G. World Map of the Status 
of Human-Induced Soil Degradation: An Explanatory Note. Wageningen: International Soil 
Reference and Information Centre (ISRIC). 34 pp. 

Oliver, M., S. Wilson, J. Gomboso, and T. Muller. 1996. The costs of salinity to government agencies 
and public utilities in the Murray-Darling basin. ABARE Research Report # 96.2. Australian 
Bureau of Agricultural and Resource Economics. 83 pp. 

Oster, J. D., and R. D. Ingvalson. 1967. In situ measurement of soil salinity with a sensor. Soil Sci. Soc. 
Amer. Proc. 31:572-574. 

Oster, J. D., and L. S. Willardson. 1971. Reliability of salinity sensors for the management of soil 
salinity. Agron. J. 63:695-698. 

Oster, J. D. and J. D. Rhoades. 1990. Steady state root zone salt balance. In Agricultural Salinity 
Assessment and Management Manual. K. K. Tanji (ed.). ASCE Manuals & Reports on 
Engineering No. 71, ASCE, New York. pp. 469-481. 

Oster, J. D., L. S. Willardson, and G.J. Hoffman. 1973. Sprinkling and ponding techniques for 
reclaiming saline soils. Trans. ASCE. 16:89-91. 

Oster, J. D., L. S. Willardson, J. van Schilfgaarde, and J. O. Goertzen. 1976. Irrigation control using 
tensiometers and salinity sensors. Trans. ASAE. 19:294-298. 

Parizek, Richard R., and Burke E. Lane. 1970. Soil-water sampling using pan and deep pressure- 
vacuum lysimeters. J. Hydrol. 11:1-21. 

Paul, J. L., K. K. Tanji and W. D. Anderson. 1966. Estimating soil sand saturation extract composition 
by a computer method. Soil Sci. Soc. Am. Proc. 30:15-17. 

Reeve, R.C., and E.J. Doering. 1965. Sampling the soil solution for salinity appraisal. Soil Sci. 99:339- 
344. 



1 24 References 



Reitemeier, R.F. 1946. Effect of moisture content on the dissolved and exchangeable ions of soils of arid 
regions. Soil Sci. 61:195-214. 

Rengasamy, P. 1997. Sodic soils. In R. Lai, W. H. Blum, C. Valentine, and B. A. Stewart (eds.), 
Methods for Assessment of Soil Degradation, CRC Press, pp. 265-277. 

Rhoades, J. D. 1972. Quality of water for irrigation. Soil Sci. 113:277-284. 

Rhoades, J. D. 1976. Measuring, mapping and monitoring field salinity and water table depths with soil 
resistance measurements. FAO Soils Bulletin. 31:159-186. 

Rhoades, J. D. 1978. Monitoring soil salinity: A review of methods. Establishment of water quality 
monitoring programs. L.G. Everett, K.D. Schmidt (eds.) Am. Water Resources Assoc, San 
Francisco, CA, June 1978. 2:150-165. 

Rhoades, J. D. 1979a. Salinity management and monitoring. Proc. 12 th Biennial Conference on Ground 
Water, Sacramento, CA, Sept. 1979. Calif. Water Resources Center. Report. 45:73-87. 

Rhoades, J. D. 1979b. Inexpensive four-electrode probe for monitoring soil salinity. Soil Sci. Soc. Am. J. 
43:817-818. 

Rhoades, J. D. 1980. Determining leaching fraction from field measurements of soil electrical 
conductivity. Agric. Water Mgt. 3:205-215. 

Rhoades, J. D. 1981. Predicting bulk soil electrical conductivity vs. saturation paste extract electrical 
conductivity calibrations from soil properties. Soil Sci. Soc. Am. J. 45:43-44. 

Rhoades, J. D. 1982. Soluble Salts. In A. L. Page, R.H. Miller and D. R. Kenney (eds.) Methods of soil 
analysis part 2, chemical and microbiological properties. Agronomy Monograph. 9:167-178. 

Rhoades, J. D. 1984. Principles and methods of monitoring soil salinity. In Soil salinity and irrigation - 
processes and management. Springer Verlag, Berlin. 5:130-142. 

Rhoades, J. D. 1990a. Sensing soil salinity problems: New technology. In R.L. Elliott (ed.) Proc. 3rd. 
National irrigation symposium irrigation association and ASCE, October 28 - November 1, 1990. 
Phoenix, AZ, pp. 422-428. 

Rhoades, J. D. 1990b. Determining soil salinity from measurements of electrical conductivity. Commun. 
Soil Sci. Plant Anal. 21:861-901. 

Rhoades, J. D. 1992a. Instrumental field methods of salinity appraisal. In G.C. Topp, W. D. Reynolds 
and R. E. Green (eds.), Advances in measurement of soil physical properties: Bringing theory into 
practice. SSSA Spec. Publ. 30. ASA, CSSA and SSSA, Madison, WL, pp. 231-248. 

Rhoades, J. D. 1992b. Recent advances in the methodology for measuring and mapping soil salinity. 
Proc. Int'l Symp. on Strategies for Utilizing Salt Affected Lands, ISSS Meeting, Bangkok, 
Thailand, Feb. 17-25, 1992. 

Rhoades, J. D. 1993. Electrical conductivity methods for measuring and mapping soil salinity. In: D. L. 
Sparks (ed.), Advances in Agonomy. Vol. 49:201-251. 

Rhoades, J. D. 1994. Soil salinity assessment: Recent advances and findings. ISSS Sub-Commission A 
Meeting, Acupulco, Mexico, July 10-16, 1994. 

Rhoades, J. D. 1996a. Salinity: Electrical conductivity and total dissolved salts. In Methods of Soil 
Analysis, Part 3, Chemical Methods, Soil Sci. Soc. Am. Book Series 5, Amer. Soc. of Agron., Inc., 
Madison, Wisconsin, pp. 417-435. 

Rhoades, J. D. 1996b. New Assessment technology for the diagnosis and control of salinity in irrigated 
lands. Proc. International Symposium on Development of Basic Technology for Sustainable 
Agriculture Under Saline Conditions, Arid Land Research Center, Tottori University, Tottori, 
Japan, December 12, 1996. pp. 1-9. 



Soil salinity assessment 1 25 



Rhoades, J. D. 1997a. Sustainability of Irrigation: An Overview of Salinity Problems and Control 
Strategies. Proc. 1997 Annual Conference of the Canadian Water Resources Association, 
Footprints of Humanity: Reflections on Fifty Years of Water Resources Developments, Lethbridge, 
Alberta, Canada, June 3-6, 1997., pp. 1-40. 

Rhoades, J. D. 1997b. Geospatial measurements of soil electrical conductivity to determine soil salinity 
and diffuse salt-loading from irrigation. Proc. Joint AGU Chapman Conference/SSSA Outreach 
Conference on the Application of GIS, Remote Sensing, Geostatistics, and Solute Transport 
Modeling to the Assessment of Nonpoint Source Pollutants in the Vadose Zone, (in press). 

Rhoades, J. D. 1998. Use of Saline and Brackish Waters for Irrigation: Implications and Role in 
Increasing Food Production, Conserving Water, Sustaining Irrigation and Controlling Soil and 
Water Degradation. In R. Ragab and G. Pearce (eds.) Proceedings of the International Workshop 
on "The use of saline and brackish water for irrigation - implications for the management of 
irrigation, drainage and crops", Bali, Indonesia, pages 23-24. 

Rhoades, J. D., and R .D. Ingvalson. 1971. Determining salinity in field soils with soil resistance 
measurements. Soil Sci. Soc. Amer. Proc. 35:54-60. 

Rhoades, J. D. and S. D. Merrill. 1976. Assessing the suitability of water for irrigation: Theoretical and 
empirical approaches. FAO Soils Bulletin 31:69-109. 

Rhoades, J. D., and J. van Schilfgaarde. 1976. An electrical conductivity probe for determining soil 
salinity. Soil Sci. Soc. Am. J. 40:647-651. 

Rhoades, J. D., and D.L. Corwin. 1981. Determining soil electrical conductivity - depth relations using 
an inductive electromagnetic soil conductivity meter. Soil Sci. Soc. Am. J. 45:255-260. 

Rhoades, J. D., and D.L. Corwin. 1984. Monitoring soil salinity. J. Soil and Water Conservation 39 (3) : 

172-175. 

Rhoades, J. D., and J. D. Oster. 1986. Solute content. In A. Klute (ed.) Methods of soil analysis, Part II. 
Physical and Mineralogical Methods 2 nd Ed. Am. Soc. Agron. Monogr. 9:985-1006, Madison, WI. 

Rhoades, J. D., and D.L. Corwin. 1990. Soil electrical conductivity: Effects of soil properties and 
application to soil salinity appraisal. Commun. Soil Sci. Plant Anal. 21:837-860. 

Rhoades, J. D., and S. Miyamoto. 1990. Testing soils for salinity and sodicity. Soil Testing and Plant 
Anal. 3 rd Ed., SSSA Book Series No. 3, pp. 299-336. 

Rhoades, J. D., J. D. Oster, R. D. Ingvalson, J. M. Tucker and M. Clark. 1974. Minimizing the salt 
burdens of irrigation drainage waters. J. Environ. Qual. 3:311-316. 

Rhoades, J. D., P. A. C. Raats, and R.J. Prather. 1976. Effects of liquid-phase electrical conductivity, 
water content, and surface conductivity on bulk soil electrical conductivity. Soil Sci. Soc. Am. J. 
40:651-655. 

Rhoades, J. D., M. T. Kaddah, A. D. Halvorson, and R.J. Prather. 1977. Establishing soil electrical 
conductivity-salinity calibrations using four-electrode cells containing undisturbed soil cores. Soil 
Sci. 123:137-141. 

Rhoades, J. D., D.L. Corwin, and G.J. Hoffman. 1981. Scheduling and controlling irrigations from 
measurements of soil electrical conductivity. Proc. ASAE Irrigation Scheduling Conference, 
Chicago, IL., Dec. 14, 1981. pp. 106-115. 

Rhoades, J. D., N. A. Manteghi, P.J. Shouse, and W.J. Alves. 1989a. Soil electrical conductivity and soil 
salinity: New formulations and calibrations. Soil Sci. Soc.Am. J. 53:433-439. 

Rhoades, J. D., N. A. Manteghi, P.J. Shouse, and W.J. Alves. 1989b. Estimating soil salinity from 
saturated soil-paste electrical conductivity. Soil Sci. Soc. Am. J. 53:428-433. 



1 26 References 



Rhoades, J. D., B.L. Waggoner, P.J. Shouse, and W.J. Alves. 1989c. Determining soil salinity from soil 
and soil-paste electrical conductivities: Sensitivity analysis of models. Soil Sci. Soc. Am. J. 
53:1368-1374. 

Rhoades, J. D., S.M. Lesch, P.J. Shouse, and W.J. Alves. 1989d. New calibrations for determining soil 
electrical conductivity - depth relations from electromagnetic measurements. Soil Sci. Soc. Am. J. 

53:74-79. 

Rhoades, J. D., P.J. Shouse, W.J. Alves, N. A. Manteghi, and S.M. Lesch. 1990a. Determining soil 
salinity from soil electrical conductivity using different models and estimates. Soil Sci. Soc. Am. J. 
54:46-54. 

Rhoades, J. D., D.L. Corwin, and S.M. Lesch. 1990b. Effect of soil EC a - depth profile pattern on 
electromagnetic induction measurements. U.S. Salinity Lab. Report. #125, 108 pp. 

Rhoades, J. D., S. M. Lesch, P. J. Shouse and W. J. Alves. 1990c. Locating sampling sites for salinity 
mapping. Soil Sci. Soc. Am. J. 54:1799-1803. 

Rhoades, J. D., A. Kandiah and A. M. Mashali. 1992. The use of saline waters for crop production. 
FAO Irrigation & Drainage Paper 48, FAO, Rome, Italy. 133 p. 

Rhoades, J. D., S. M. Lesch, R. D. LeMert and W. J. Alves. 1997a. Assessing 
irrigation/drainage/salinity management using spatially referenced salinity measurements. Agr. 
Water Mgt. 35:147-165. 

Rhoades, J. D., S. M. Lesch, S. L. Burch, R. D. LeMert, P. J. Shouse, J. D. Oster and T. O'Halloran. 
1997b. Salt transport in cracking soils: Salt distributions in soil profiles and fields and salt pick-up 
by run-off waters. J. Irrig. & Drainage Engineering, ASCE 123:323-328. 

Rhoades, J. D., N. A. Manteghi, S. M. Lesch and D. C. Slovacek. 1997c. Determining soil & water 
sodicity from electrode measurements. Commun. Soil Sci. Plant Anal. 28:1737-1765. 

Rhoades, J. D., S. M. Lesch and D. B. Jaynes. 1997d. Geospatial measurements of soil electrical 
conductivity for prescription farming. Special Publ. Soil Sci. Soc. Am., ASA, Madison, WI. (in 
press). 

Richards, L. A. 1941. A pressure-membrane extraction apparatus for soil solution. Soil Sci. Soc. 51:377- 
386. 

Richards, L. A. 1966. A soil salinity sensor of improved design. Soil Sci. Soc. Amer. Proc. 30:333-337. 

Rose, C. W., P. W. A. Dayananda, D. R. Nielsen and J. M. Biggar. 1979. Long-term solute dynamics 
and hydrology in irrigated slowly permeable soils. Irrig. Sci. 1:77-87. 

Ross, D.S., and R.J. Bartlett. 1990. Effects of extraction methods and sample storage on properties of 
solutions obtained from forested spodosols. J. Environ. Qual. 19: 108-1 13. 

Seckler, David. 1996. The New Era of Water Resources Management: From "Dry" to "Wet" Water 
Savings, Consultative Group on International Agricultural Research, Washington, D C. 

Shainberg, I., J. D. Rhoades, and R. J. Prather. 1980. Effect of exchangeable sodium percentage, cation 
exchange capacity, and soil solution composition on soil electrical conductivity. Soil Sci. Soc. Am. 
j. 44:469-473. 

Shaw, R. 1994. Estimation of the electrical conductivity of saturation extracts from the electrical 
conductivity of 1:5 soil: water suspensions and various soil properties. Project Report Series 
QO94025, Dept. of Primary Industries, Queensland, Australia, 41 pp. 

Shimshi, Daniel. 1966. Use of ceramic points for the sampling of soil solution. Soil Sci. 101:98-103. 

Slavich, P.G. and B. J. Read. 1983. Assessment of electromagnetic induction measurements using an 
inductive electromagnetic soil conductivity meter. Soil Sci. Soc. Amer. J. 45:255-260. 



Soil salinity assessment 1 27 



Slavich, P.G. 1990. Determining EC a - depth profiles from electromagnetic induction measurements. 
Aust. J. Soil Research 28:443-452. 

Slavich, P.G. and J. Yang. 1990. Estimation of field leaching rates from chloride mass balance and 
electromagnetic induction measurements. Irrig. Sci. 11:7-14. 

Slavich, P. G. and GH. Peterson. 1990. Estimating average rootzone salinity from electromagnetic 
induction (EM-38) measurements. Aust. J. Soil Res. 28:453-463. 

Soil Science Society of America. 1996. Methods of Soil Analysis, Part 3, Chemical Methods. American 
Society of Agronomy, Madison, Wisconsin, USA. pp. 1389. 

Sonnevelt, C, and J. van den Ende. 1971. Soil analysis by means of a 1:2 volume extract. Plant Soil. 
35:505-516. 

Suarez, D.L. 1986. A soil water extractor that minimizes CO2 degassing and pH errors. Water Resour. 
Res. 22:876-880. 

Suarez, D.L. 1987. Prediction of pH errors in soil water extractors due to degassing. Soil Sci. Soc. Am. 
151:64-67. 

SURFER. 1986. Reference manual. Golden Software, Inc., Golden, CO. 

Tadros, V. T., and J.W. McGarity. 1976. A method for collecting soil percolate and soil solution in the 
field. Plant and Soil. 44:655-667. 

US Salinity Laboratory Staff. 1954. Diagnosis and improvement of saline and alkali soils. USDA 
Handbook 60, U.S. Government Printing Office, Washington, D. C. 

van De Pol, R.M., P.J. Wierenga, and D. R. Nielsen. 1977. Solute movement in a field. Soil Sci. Soc. 
Am. J. 41:10-13. 

van Hoorn, J.W. 1980. The calibration of four-electrode soil conductivity measurements for determining 
soil salinity. Proc. Int. Symp. Salt Affected Soils, Karnal, pp. 148-156. 

Wagner, George H. 1965. Changes in nitrate N in field plot profiles as measured by the porous cup 
technique. Soil Sci. 100:397-402. 

Webster, R. 1985. Quantitative spatial analysis of soil in the field. Advances in Soil Science. 31:505- 
524. 

Webster, R. 1989. Recent achievements in geostatistical analysis of soil. Agrokemia Es Talajtan. 
38:519-536. 

Wenner, F. 1916. A method of measuring earth resistivity. U. S. Dept. Com. Cur. Of Stand. Sci. Paper 

No. 258. 

Wesseling, J., and J. D. Oster. 1973. Response of salinity sensors to rapidly changing salinity. Soil Sci. 
Soc. Amer. Proc. 37:553-557. 

Wilcox, L.V. 1951. A method for calculating the saturation percentage from the weight of a known 
volume of saturated soil paste. Soil Sci. 72:233-237. 

Williams, B. G. and G. C. Baker. 1982. An electromagnetic induction technique for reconnaissance 
surveys of soil salinity hazards. Aust. J. Soil Research 20: 107-1 18. 

Wollenhaupt, N. C, JL. Richardson, J.E. Foss, and E. C. Doll. 1986. A rapid method for estimating 
weighted soil salinity from apparent soil electrical conductivity with an above ground 
electromagnetic induction meter. Can. J. Soil Sci. 66:315-321. 

Wolt, J., and J.G. Graveel. 1986. A rapid method for obtaining soil solution using vacuum displacement. 
Soil Sci. Soc. Am. J. 50:602-605. 

Wood, J. D. 1978. Calibration stability and response time for salinity sensors. Soil Sci. Soc. Am. J. 
42:248-250. 



1 28 References 



Wood, Warren W. 1973. A technique using porous cups for water sampling at any depth in the 
unsaturated zone. Water Resour. Res. 9:486-488. 

Yadav, B. R., N. H. Rao, K.V. Paliwal, and P. B. S. Sarma. 1979. Comparison of different methods for 
measuring soil salinity under field conditions. Soil Sci. 127:335-339. 

Yamasaki, S., and A. Kishita. 1972. Studies on soil solution with reference to nutrient availability. I. 
Effect of various potassium fertilizer on its behavior in the soil solution. Soil. Sci. and Plant Nutr. 
18:1-6. 



Soil salinity assessment 1 29 



Annex 1 

Methods for establishing EC e = F(Ec a ) 

calibrations 



EC e -EC a calibrations may be directly established in one of four ways depending on equipment 
availability, time availability, and desired accuracy. 

1. The earliest-used, simplest and least accurate method is to measure EC a (using a surface 
array of four-electrodes or EM-38 sensor) at numerous locations in the field (or area) of 
interest, to analyze for salinity (EC e ) soil samples collected at no less than 4-6 sites that 
provide an approximately equally spaced interval of EC a readings over the observed range 
and to determine the best-fit linear relation between the EC e and EC a data-pairs using a 
graphical plot or standard regression analysis procedures. Examples of calibrations 
established in this manner are shown in Figures 48 and 49. Since soil salinity is typically 
quite variable from spot-to spot and with depth in saline fields/areas, numerous samples 
should be taken within the dimensions of the measurement- volume given in Chapter 3 for the 
four-electrode and EM-38 sensors, respectively, in order to obtain a representative sample of 
the relatively large volumes of soil measured by these sensors. For any site, the samples 
should be composited for each depth-increment of interest. This type of calibration is limited 
to whatever range of salinity exists in the field (area) at the time of sampling. The accuracy 
is limited by the variability in soil properties that exist within the sampled area and by the 
degree to which the relatively small soil sample represents the much larger volume of soil 
measured by the sensors. The accuracy is generally sufficient for salinity diagnosis and gross 
mapping purposes, but not for certain other assessment purposes. 

2. A more accurate method is depicted in Figure 17. In this method a soil EC-probe is used to 
determine the EC a values of small volumes of soil that have been adjusted in the field to 
provide a desired optimum range and distribution of salinity values. To adjust the salinity, 
saline waters of various salinities (EC = 5, 10, 20, 40, etc.; SAR values = 8, or whatever else 
is deemed appropriate) are impounded in column sections (30 cm in diameter by 45 cm in 
length) driven about 10-15 cm into the soil and in a surrounding 15 -cm wide excavated moat 
(see Figure 17a). (About 40 litres of saline water is required to bring the soil to a depth of 30 
cm beneath the impounded area to the desired level of salinity). 

When the soil has drained to about "field-capacity" water content, 2-3 days after the 
impounded water has infiltrated, a 2.5 cm hole is cored to a depth of 30 cm in the centre of 
the uniformly salinized soil volume using a Lord, or equivalent diameter coring tube (see 
Figure 17b). Next, the soil EC-probe is inserted (see Figure 17c) into the slightly undersized 
hole to a depth that centres its electrodes at the desired depth (usually 15 cm) and the reading 
of EC a is obtained along with the temperature (using either the thermistor in the Martek 



130 



Annex 1: Methods for establishing EC e = F(EC„) calibrations 



FIGURE 1.1 

Series of four-electrode cells containing 
undisturbed soil-cores being segmented after 
removal from the soil-core sampler 



probe or any suitable temperature 
probe). After the soil EC -probe is 
removed, the soil immediately exterior 
to the position of the centered EC- 
probe is sampled (essentially the 7.5- 
22.5 cm depth) using a 10-15 cm 
diameter soil auger (see Figure 17d). 
This sample of soil (which closely 
correspond to the volume measured by 
the EC-probe for EC a ) is analysed by 
any conventional method for salinity 
(EC e ). The linear EC e -EC a calibration 
is established by graphical means or by 
regression analysis. Example 

calibrations of this type are shown in 
Figures 19 and 20. Such calibrations 
are the quickest to obtain of any of the 
methods and are generally quite 
accurate because the volumes of soil 
sampled for EC e and EC a are nearly the 
same and because the variability in soil 
type and water content is minimized, as 
is salinity (within the sampled region of 
soil). Such calibrations have been used 
to establish predictive calibration- 
relations by soil type using auxiliary 
soil-property data of the soil samples 

obtained during the calibrations of numerous different soils. Very nice relations have been 
established in this regard as previously discussed in the main text sections of this report. 





M^^^W^^f 




• 








i. 

t . 


M 






r - 


y 


•/ Wlm ■ 




" 



3. A still more accurate direct calibration method has been developed using specially built four- 
electrode cells that fit as removable inserts within soil coring devices used to obtain 
undisturbed soil-core samples. Undisturbed soil cores are collected from the soil sites 
representative of the soil type, field, or area of interest (or from the soil bodies whose salinity 
has been adjusted by the previously described method) using lucite column sections as corer- 
inserts.. A soil-filled four-electrode cell is obtained by slicing through the soil core, removed 
from the corer, between adjacent cylinder segments and then screwing electrodes into the 
tapped hoes in the cell wall, as illustrated in Figures 1 8 and 1.1. The composite EC a of the 
soil is calculated from the succession of readings made by sequentially connecting the four- 
electrode generator/meter to the eight combinations of electrodes that can be achieved with 
the eight equidistant-spaced electrodes inserted into the cell/soil (only four electrodes are 
shown in Figure 18). The cell constant (k) of these four-electrode cells are obtained by filling 
them with standard EC-solutions and measuring the resistance of the filled-cell using 
Equations [1] to [3]. The soil temperature is also measured using a suitable temperature 
probe and EC a is calculated from the these data using Equations [1] to [3]. After EC a is 
measured, the soil is removed from the cell and analyzed for salinity (and other properties of 
interest). Next, the linear EC e -EC a calibration is established as described for the other 
methods. These types of calibrations are very accurate because the measurements of EC e and 
EC a are made on exactly the same volume of soil. An example of the improvement in 



Soil salinity assessment 



131 



accuracy is illustrated in Figure 1-2; 
the improvement in the lower range of 
salinity is especially evident. The 
above-described methods of 

calibration are described in more 
detail elsewhere (Rhoades, 1976; 
Rhoades and Halvorson, 1977; 
Rhoades et al, 1977). 

4. An analogous but more accurate and 
statistically rigorous method for 
establishing EC e -EC a calibrations that 
apply to specific field situations is the 
stochastic, field-calibration method 
previously described in the main text. 



FIGURE 1.2 

Comparison of EC e -EC a calibrations as 

determined by methods (1) and (3) 



35 



I 






* f IttlrKli Oil 




I t 3 4 a * f ■':•"■ .* « 



132 Annex 1: Methods for establishing EC e = F(EC„) calibrations 



Soil salinity assessment 



133 



Annex 2 



Circuitry and parts-list for soil EC-meter 



The circuit developed for use with soil EC-probes is shown in Figures 2.1 and 2.2. The 
components cost is less than about US$100. With this circuit, a current is passed through the 
outside pair of current electrodes, and the voltage across the inside pair is measured. Since, the 
current and voltage are known, the resistance can be calculated and, in turn, the conductivity of 
the soil determined using Equation [24]. The accuracy of the circuit is ± 0.3 ohms from 0-100 
ohms and ± 4 ohms from 100-200 ohms (or 2 % of range). For more detail about the operation 
of the circuit see Austin and Rhoades (1979). 



FIG URE 2.1 

Diagram of a sim plified circuit or E C a -meter; R y , the resistance of the null adjustment 
potentiometer, is adjusted to equal R,, the resistance of the salinity sensor (after Austin and 
Rhoades, 1979) 



CURRENT LIMITING RESISTOR 

i $ r 



COMPARATOR 




LEO"* 



134 



Annex 2: Circuitry and parts-list for soil EC-meter 



FIGURE 2.2 

Detailed diagram of low-cost circuit for reading four-electrode sensors (after Austin and 

Rhoades, 1979) 



114 ■ HI Hktl-I- 



t-.tsy Ft 







1 II 1" 




Soil salinity assessment 1 35 



Annex 3 

Equation for calculating effect of insertion- 
depth of four-electrodes 



The depth of electrode-insertion affects the measurement of soil electrical conductivity. For 
electrodes configured equidistantly apart in the so-called Wenner-array, the relationship between 
EC a , distance between electrodes (a, in m) and depth of electrode insertion (b, in m) is, according 
to Wenner (1916), as follows: 



f \ 



EC„ 



1+ 



1+4 {b/af ^l+{b/af 



2 
v » » > - j 



'{4%aR), [43] 



where R is resistance in ohms. The term inside the brackets in the numerator approaches the 
limiting value of 2 as the depth of electrode insertion (b) becomes small relative to the distance 
between the electrodes (a). 



136 Annex 3: Equation for calculating effect of insertion depth of four-electrodes 



Soil salinity assessment 



137 



Annex 4 

Construction of burial-type 
four-electrode probe 



Salinity monitoring sometimes requires that repeated measurements be made over a period of time 
at the same location. For such uses, implanted probes offer certain advantages. For this reason, 
an inexpensive four-electrode unit was developed that can be implanted and left in the soil for 
extended periods of time. This burial-type probe is constructed using the components shown in 
Figure 4. 1 . 



FIG U RE 4,1 

In ex p en s ive four-electrode burial-type p ro be be to re and after a s s e m b ly (after Rhoades, 1979) 




The probe casing is a 4 3/4 inch long length of PVC pipe (3/4 inch schedule 80) in which 
four grooves (0.040 inch wide and 0.025 inch deep) are made. The distance between the outside 
grooves is 2.5 inches; the inside grooves are inset by 0.25 inches. The outside wall of the pipe is 
tapered at 0.5° on a lathe. Two holes (1/16 inch in diameter) are drilled in each of the grooves for 
the passage of 18-gauge thermostat-type wire, with the insulation removed from the last 4 inches. 
Each of the four bare wire-ends is passed from inside of the pipe through a hole, is rapped around 



138 Annex 4: Construction of burial-type four-electrode probe 



the outside of the pipe in one of the grooves to form one of the four electrodes, and is returned to 
the inside of the pipe through a second hole in each groove. The returned end of the wire is 
crimped against the inside pipe wall to firmly secure the wire in place. A 1.37 inch- long tapered 
(7° on a side) solid PVC tip is inserted into the head-end of the probe and cemented in-place with 
PVC solvent cement to serve as a leading edge. Finally, the inside of the pipe is filled with 
laminating resin to within 0.75 inches of the exit end in order to prevent water entry. The 
materials cost is less than about US$1. per probe. Data illustrating the utility of the burial-probe 
are given in Figures 80, 81, and 82. More details about the construction of these probes and 
about devices to facilitate their installation in the soil are given in PJioades (1979). The 
commercialized version of the probe, which consists in part of plastic and which is formed by a 
pressure injection/molding techniques, is shown in Figure 38. 



Soil salinity assessment 



139 



Annex 5 

Examples of various special-use 
four-electrode cells and sensors 



FIG U RE 5,1 

Schematic of apparatus used to vary water content 
and co m p o s itio n and o I a four-electrode cell used to 
measure the associated values of soil electrical 
conductivity (after Bottraud and Rhoades, 1 985b) 



Various types of four-electrode cells and miniaturized four-electrode sensors have been developed 

for special needs. A few are shown here to illustrate how easy they are to make and to provide 

ideas for their construction, as well as for others. The detailed information is not provided, since 

the exact dimensions vary with 

purpose and availability of 

materials. Figures 5.1 and 5.2 

illustrate a four-electrode cell that 

permits the composition and water 

content in soil cores (either packed 

or undisturbed cores) to be varied 

and the corresponding value of soil 

electrical conductivity to be 

determined (after Bottraud and 

Rhoades, 1985b). Two other 

versions of cells used to adjust soil 

columns to different solution 

compositions, with and without the 

opportunity for changes in soil 

volume, are shown in Figure 5.3 

(after Bottraud and Rhoades, 

1985a). Analogous versions shown 

in Figure 5.4 have been made with 

ceramic plates as bases to permit 

the application of pressure so as to 

be able to vary water content over a 

greater range than possible with the 

cells shown in Figures 5.1 and 5.2. 

Even greater pressures can be 

applied to vary water content using pressure plate apparatus and the types of cells and 

undisturbed soil cores. Micro four-electrode sensors, such as those shown in Figures 5.5 and 5.6, 

can be constructed to permit the measurement on EC a in very shallow soil depths (Figure 5.6) or 

in small increments of an exposed (excavated) soil profile (Figure 5.5). 



dir 




sflluiion conduction y 

Mil" 

glass tubtrq 

\ lygon tut)ing<: 

1 — T '■ ' ' "T f"*^ 



■wlulion 



\J 



140 



Annex 5: Examples of various special-use four-electrode cells and sensors 



FIGURE 5.2 

Several kinds of four-electrode cells and bases used to vary water content and composition 

and to measure the associated values of soil electrical conductivity 




FIGURE 5.3 

Schematic of apparatus used to hold four-electrode cells for studying the effects of varying 
solution composition and induced changes in water content and porosity and to measure 
the associated values of soil electrical conductivity (after Bottraud and Rhoades, 1985a) 




\ 



\ 



wotef 
soil — 



HP- 




Mm 




Constrained Soil 



Unconstrained Soil 



Soil salinity assessment 



141 



FIGURE 5.4 

An apparatus built to hold a four-electrode 
cell with ceramic containing end-plates to 
permit changes in water content to be 
induced by the application of pressure 




FIGURE 5.5 

Small-span four-electrode sensor used to 
measure soil electrical conductivity in 
small depth-increments along an exposed 
soil profile 




142 



Annex 5: Examples of various special-use four-electrode cells and sensors 



FIGURE 5.6 

(A) Small-span four-electrode array used to measure soil electrical conductivity in shallow 

depths, and (B) connection of meter to four-electrode cell (after Rhoades ef a/., 1977) 




Soil salinity assessment 1 43 



Annex 6 



Derivation of EM a dj equations 



This appendix gives the original basis of the relations of the type given in Equation [22] and 
Tab lei for estimating the EC a levels for different depth-intervals of the soil profile using only two 
EM-38 readings. In this approach the EM-38 readings are related to EC, by a series of simple 
equations which are based, in part, upon the theoretical response functions of the sensor for 
homogeneous media as given in Figure 44. This figure shows that EM H and EM V measurements 
give depth-weighted EC a values to about 1 and 2 metres, respectively. For the to 0.3-m 
increment of soil, the following relations apply for homogeneous profiles: 

EM„, V =0.150 EC -o.3,v +0.850 EC >0 .3,v and [a] 

EM„.h = 0.435 ECo-o.3, h + 0.565 EC >0 . 3 , „ [b] 

where EM 0jV and EM . H are the EM-38 values measured at the soil surface in the vertical and 
horizontal positions, respectively; and EC0-0.3, v , EC >0 .3,v , EC0-0.3, h , and EC> .3, h are the actual 
EC a values for the to 0.3-m and >0.3-m soil depth intervals. He subscript ", " in EC a is dropped 
from these equations and some of the following ones in order to minimize the clutter in the 
subscripts. In an homogeneous profile, the to 0.3-m depth of soil only contributes 15 % of the 
EM 0iV value, while the deeper depths contribute 85 %. The corresponding values for the EM 0>H 
value are 43.5 % and 56.5 %, respectively. 

Since the volume of soil measured within the to 0.3-m increment is very similar for the 
vertical and horizontal orientations, it is reasonable to assume that EC0-0.3, v = EC0-0.3, h • However, 
in the case of the >0.3-m increment, the volumes of measurement are quite different and, 
consequently, EC> 3 .\ and EC >0 3. h can not be assumed to be equal in value. However, in order to 
establish a relationship between EC a 0-0.3, EM 0?V and EM . H using equations [a] and [b], it is 
necessary to equate EC >0 .3,v and EC> .3, h • This problem was overcome when it was found using 
empirical data that EM .h could be adjusted using empirical relationships (see Figure 6.1) so that 
EC>o.3,v calculated from Equation [a] would equal EC >0 .3, h calculated from Equation [b]. The 
empirical data and its means of collection are explained in detail in Corwin and Rhoades (1982, 
1990). Briefly, an adjusted EM 0>H (for the to 0.3-m increment) was calculated from Equation 
[b] using the measured values of EC,, 0-0.3 and the values of EC> 3 ,v calculated from Equation [a]. 
The plot of measured and adjusted EM 0>H values for each depth increment of the test soils 
revealed the set of linear relations shown in Figure 6.1. Assuming these relations would apply to 
other soils, measured values of EM», H for a specified depth increment (0-h metres) can be adjusted 
so that EC>l v = EC> h , h , as was demonstrated for the to 0.3-m depth-increment as follows: 



144 



Annex 6: Derivation ofEM aii j equations 



FIGURE 6.1 

Relationship between electromagnetic soil conductivity as measured by the EM-38 in the 
horizontal position at the soil surface, EM , h (measured) and adjusted electromagnetic soil 
conductivity, EM ,h (adjusted) for composite depths (after Corwin and Rhoades, 1990) 

^justed EI«!romopi?lK Conduelmry CiKvti l« CompOWJe Increment* 



Ei 







I 2 1 * 5 « 7 

rfAijfttD £ l£C ThMM Tit ijuujl.- VITT 




k J i * S t 7 1 9 D 

i**sui*p (LfcnrauicT*: ooHDuctivrrf 




o i J } 4 * * I i 9i 



Hi 



II 

m 
i 

1 I 











n-i?« y 












if 


















Mm 






jT Him- TtHmmmtfWM 


■ 




/• *. ..■■■*» 





* « 



i * id ii 



■CttiJWD tLrCTFKtUBCTIC CDUKTMTV 



EM ,v = 0.150 ECo-o.3 + 0.850 EC >0 3 ,v and [c] 

EM, |H (adjusted, 0-0.3-m) = 0.435 EC()-0.3 + 0.565 EC>o,3. v . [d] 

Equations [c] and [d] can now be reduced by substitution to form the following single 



equation: 

EC a , 0-0.3 = 2.982 EM 0j h (adjusted, 0-0.3-m) - 1-982 EM 0>V , [e] 

where EM .H (adjusted, 0-0.3-m) is an empirically obtained linear expression of the form shown in Figure 
94, such as: 

EMo.H (adjusted, 0-0.3-m) = kl EMo,H + <k • [f] 

Equations [e] and [f] can be reduced by substitution to form a single equation. Following 
the same rationale, an analogous set of equations can be obtained to predict EC a for other soil 
depth intervals from EM .v and EM U . These equations are of the form: 



EC, 



= ks EM ,H - k v EM ,v + k , 



[g] 



where k) , k v , and k are empirically determined coefficients for the depth interval Xi -x 2 . The value 
of k should ideally be zero, but often is not, due to experimental error in the data. Equation [22] 
is a modification that was undertaken to achieve a more normal distribution of data so as to 
establish better values of the coefficients by statistical methods. Equation [23] is a further 
modification that was undertaken for the reasons explained in the main text. 



Soil salinity assessment 



145 



Annex 7 

Device for positioning EM-38 sensor during 

hand-held measurements 



The device used to position the midpoints of the EM-38 magnets at a height of 10-cm and 50-cm above 
the ground when readings are taken in either the vertical or horizontal configurations is shown in Figure 
(7.1 A and 7. IB) and Figure (7.2A and 7.2B), respectively. It is a block of lightweight wood (redwood) 
of dimensions 8.5 by 8.5 by 50 cm in which a 3.5 cm wide groove is cut 4.7 cm deep along the 
longitudinal axis and 6.0 cm deep in the end section (see Figure 7.3). A handle is attached by wooden 
dowels, plastic screws and glue; metal is not used or contained in any part of the "block". The EM-38 is 
laid on its side for the EM H reading and is placed in the slot for the EM V reading. 



FIG URE 7.1 

EM-38 sensor (centre of coils) positioned 10 

h o rlzo n tal o rlen tatlo n s, respectively 



cm above ground In the (A) vertical and (B 



146 Annex 7: Device for positioning EM- 3 8 sensor during hand-held measurements 



FIGURE 7.2 

EM-38 sensor (centre of coils) positioned 50 cm above ground in the (A) vertical and (B) 

horizontal orientations, respectively 





FIGURE 7.3 

Wooden device used to position EM-38 at 10- and 50-cm heights above ground 




Soil salinity assessment 



147 



Annex 8 

Schematic and parts-list for soil four-electrode 

probe 



FIG URE 8.1 

Schematic of the design and parts list for the construction 
of a four-electrode soil EC-probe (after Rhoades and van 
Schilfgaarde, 1976) 



USS.L SALINITY PR06E 



c 



r tm . 


1- 





* 




fcojwur 


m 


! 


^ MtJ - M 


t 


" 


lhK*hu < ' 


P 


■ 


h.« 


f 


1 


l-V-M- 


r . «■— * -^- 


■ 




■nd. 


H 






1 I - "*'- 1- j'n 

Pi— i *- 



The construction details of a 
mechanically fabricated soil 
EC -probe are provided in 
Figure 8- 1 . Four brass annular 
rings (electrodes) are juxta- 
posed between lucite 
insulators/spacers to form the 
probe. Teflon gaskets are 
placed between the electrodes 
and insulators/spacers, along 
with epoxy sealer, to prevent 
water from entering and 
shorting out the electrodes (see 
Figure 8. 2 A). The size of the 
probe is dimensioned to permit 
EC a to be measured in 15 -cm 
increments. The probe is 
afixed to a thick-walled, 
anodized aluminum shaft so 
that it can be inserted to the 
desired depth in the soil via a 
hole made with a standard 2.3 
cm Oakfield or Lord soil 
sampler (see Figure 8.2B). 
The probe is slightly tapered 
(1°) toward the tip so that all 
four electrodes firmly contact 
the soil upon insertion in the 

hole. The leads from the electrodes exit the handle for connection to the generator/meter. More 
details and description of this soil EC-probe is given in Rhoades and van Schilfgaarde (1976). 
The commercial probe sold by the Eijkelkamp Agrisearch Equipment Company is essentially a 
copy of the above-described unit; that sold by Martek Instruments is a plastic-moulded version. 







148 



Annex 8: Schematic and parts-list for soil four-electrode probe 



FIGURE 8.2 

(A) An unassembled and (B) assembled four-electrode soil EC-probe 




Soil salinity assessment 1 49 



Annex 9 

Description of statistical tests for monitoring 

soil salinity 



The periodic assessment of soil salinity conditions over time is a critical component of any sort of 
serious, long-term monitoring strategy. The selection and acquisition of soil samples should be 
performed in a manner which optimizes the possibility of detecting any temporal and/or spatial- 
temporal trends occurring in the field. Lesch et al., 1998, describes a regression based testing 
methodology that can be used to detect such trends. In this approach, and electromagnetic (EM) 
survey is performed at grid-points across the field and a limited number of soil salinity samples 
are then acquired at a selected small number of these survey grid sites. An EM-salinity regression 
model (Equation [34]) is then estimated and used to calculate (i.e., predict) the soil salinity levels 
at every grid site. Once this model has been estimated, new salinity samples can be acquired in 
the future at one or more of the known survey grid sites and compared to the predicted salinity 
levels (from the model). Lesch et al., 1998, show that two statistical tests can be developed from 
such data: (1) a test to detect dynamic spatial variation over time - i.e., has the pattern changed in 
a non-random, dynamic manner across the field, and (2) a test to detect a shift in the median field 
salinity level - i.e., has the median salinity level increased up or decreased over time. This 
methodology does not require extensive soil sampling. For example, 12 to 20 soil samples are 
usually sufficient for establishing the initial regression equation, and the periodic acquisition of 8 
tolO new soil samples are typically sufficient for testing purposes. 

In more formal terms, the above tests are based on fitting a conditional regression model to 
the EM-salinity data, where the EM covariate data is assumed to be random (i.e., implying that 
the regression equation is really a mixed linear model). To do so, define y^ and y 2k as the 
observed In salinity levels from the j th and k th sample site acquired during the first and second time 
frames, where j = 1, 2, ..., n and k = 1, 2, ..., m. Let yi represent the vector of observations from 
the first time frame, and y 2 represent the observations from the second time frame. Additionally, 
define X| as the matrix (grid) of EM covariate signal data observed during the first time frame. 
For this discussion, suppose that a survey grid of size N (N > n,m) of representative EM 
covariate data has been acquired during the first time frame only, and that the n and m sample 
sites (from the first and second time frames, respectively) are chosen from this grid. Note that the 
two sets of sample sites need not be collocated. Furthermore, assume that the conditional 
distribution of y given the observed X| matrix is Normal with constant variance, and that the 
distribution of X| does not depend on either the (3 parameter vector or the error variance. Assume 
that a suitable model for the first time frame is: 

y, IX, = x,p + ei (a) 



150 Annex 9: Description of statistical tests for monitoring soil salinity 



where £| ~ N(0,G*I), and I represents the identity matrix. Furthermore, assume that a suitable 
model for the second time frame is: 

y: IX, = X,p + d„+Ti + G 2 (b) 

where d = [do, do,... ,do], r\ ~ N(0,9 2 I), e 2 ~ N(0,<Xl), and the T|, g,, and e 2 random error 
components are mutually independent. Hence, (y 2 I X L ) - (y, I X,) = d + r\ + e 2 - Gi, where do 
represents the shift in the average In salinity level between the two time frames and the additional 
error term r\ represents the dynamic variability component. 

Now, suppose that m new samples located on the grid are acquired during the second time 
frame. Let y 2 represent this vector of sample observations, y, represent the corresponding vector 
of predicted levels computed from equation (1) at these m sites, and define H m = X m T (X„ T X n ) ' X m , 
where X m represents the matrix of EM covariate data associated with these m prediction sites 
from the first time frame. Then equation (2) implies that the prediction error associated with 
these sites would be (y 2 -y) IX, ~ N( d , 6 2 I m +0 2 (I m + H m )). Now, let d = y 2 -y, where = {d, d 2 , 
..., d m }. Define the calculated sample mean and variance of these observed differences as u and 
w 2 , where u = (l/m)[d+d+...+d m ] and w 2 = (l/(m-l))[(d-u) 2 +(d-u) 2 +...+(d m -u) 2 ]. Clearly u 
represents a conditionally unbiased estimate of do. Furthermore, given the previously stated 
modelling assumptions, the following three results can be derived: 1) an F-test for determining if 
8 > 0, 2) a method of moments estimate of 0" , and 3) an approximate t-test for determining if d> 
= 0. These results are given below: 

1 . An F-test for determining if 2 > can be computed as (]) = (d-u) T E" 1 (d-u) / (m- l)s 2 , where E 
= (I + H m ), and where ()) is compared to an F distribution with m-\ and n-p-l degrees of 
freedom. 

2. The expected value of w is 6" + cf(l+A,i-A, 2 ), with X\ = (l/m)E hi and A, 2 = (\l(m(m-\)))Y!L 
hj V i^j (where hj represent the i th ,j th diagonal element of the H m matrix). Hence, a method 
of moments estimate of 0" is v = w - s"(l+X,i-X, 2 ). 

3. An approximate t-test for do = can be computed as c = u / g, where g" = (l/m)v* + 
2s 2 [(l/m) + h m „ ], h mil = Xn/CX/Xj'xn,,, x mu = (l/m)[x, + x 2 + ... + x m ], and where c is 
compared to a t distribution with n-p-l degrees of freedom. This test statistic assumes that 
the two sets of soil samples are not collocated. 

Note that the F-test represents a test for dynamic salinity variation, while the t-test represents a 
test for a shift in the median level over time.